October: critical PHP flaws need patching; Gemini 4 Argon reached defenders before developers

Every numbered sentence above rests on the facts listed under its number. A carried fact was checked when its story was published; it says when, and we did not open its source again for this piece.

1. PHP news archive, 2026 · PHP Group · archived copy

   1. 1 The PHP project published a security advisory and four releases on 21 October 2026: PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35.
      Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
      21 Oct 2026 Security advisory: Critical remote code execution vulnerabilities, update now PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35 fix two critical vulnerabilities that can allow unauthenticated remote code execution.
   2. 1 The PHP project’s advisory of 21 October 2026 lists those at risk as "all PHP installations serving web requests, including PHP-FPM, Apache with mod_php, php-cgi, FrankenPHP and LiteSpeed".
      Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
      Who is at risk: all PHP installations serving web requests, including PHP-FPM, Apache with mod_php, php-cgi, FrankenPHP and LiteSpeed.
   3. 11 The PHP project's advisory of 21 October 2026 says PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35 fix two critical vulnerabilities that can allow unauthenticated remote code execution.
      Carriedfrom Stamos says attackers exploited PHP flaws before the fix; PHP urges updating now · checked 26 Oct 2026, not opened again for this piece
      21 Oct 2026 Security advisory: "PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35 fix two critical vulnerabilities that can allow unauthenticated remote code execution."
   4. 14 The 23 October notice says: "We suggest you set session.upload_progress.enabled=0 until the next regular bugfix release, to fully protect from an unlikely RCE vector."
      Carriedfrom PHP suggests disabling session upload progress; the flaw needs a runtime serialize-handler switch. · checked 26 Oct 2026, not opened again for this piece
      We suggest you set `session.upload_progress.enabled=0` until the next regular bugfix release, to fully protect from an unlikely RCE vector.
   5. 14 The notice says: "Since this setup is unlikely in practice, we are handling it in the regular release cycle and will publish a full advisory for CVE-2026-115560 with the next release."
      Carriedfrom PHP suggests disabling session upload progress; the flaw needs a runtime serialize-handler switch. · checked 26 Oct 2026, not opened again for this piece
      Since this setup is unlikely in practice, we are handling it in the regular release cycle and will publish a full advisory for CVE-2026-115560 with the next release.
   6. 15 PHP's 2026 news archive, opened on 5 November 2026, lists no PHP release after the 8.5.12, 8.4.27, 8.3.36 and 8.2.35 releases of 21 October other than updated Windows builds of 26 October that use the same source releases, and no full advisory for CVE-2026-115560.
      checked 5 Nov 2026
      26 Oct 2026 Updated PHP Windows builds for Brotli security fix ... These builds use the same PHP source releases, and the PHP runtime versions remain unchanged. 23 Oct 2026 Informational advisory: Session upload progress ... will publish a full advisory for CVE-2026-115560 with the next release. 21 Oct 2026 Security advisory: Critical remote code execution vulnerabilities, update now
   7. 23 PHP lists mitigations for both flaws (such as session.upload_progress.enabled=0 for CVE-2026-109619) and says they do not make an installation safe.
      Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
      while only an update is a complete solution, the following measures reduce exposure to specific attack paths but do not make an installation safe. ... Set session.upload_progress.enabled=0 and restart every worker.

2. Coordinated disclosure of critical vulnerabilities (Patch the Planet) · OpenAI · archived copy

   1. 2 OpenAI says Calif identified a critical flaw using GPT-5.6 Sol as part of OpenAI's Patch the Planet initiative, and that Wiz independently discovered and reported both vulnerabilities.
      Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
      Calif identified a critical flaw using GPT‑5.6 Sol as part of our Patch the Planet initiative, with Wiz also independently discovering and reporting both vulnerabilities.

3. PHP2Shell: AI uncovers critical RCE vulnerabilities affecting 70% of the Internet · Wiz · archived copy

   1. 2 Wiz says: "We used Wiz Atlas, Gemini 4 Argon and GPT-5.6 Sol throughout this research."
      Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
      We used Wiz Atlas, Gemini 4 Argon and GPT-5.6 Sol throughout this research.
   2. 3 Wiz says: "With frontier AI models and the right tools, a single researcher completed in a week what previously would have taken a world-class team months of work."
      Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
      With frontier AI models and the right tools, a single researcher completed in a week what previously would have taken a world-class team months of work.

4. Gemini 4 Argon: our next era of frontier intelligence · Google · archived copy

   1. 4 Google published Gemini 4 Argon: our next era of frontier intelligence on 30 September 2026.
      Carriedfrom Google releases Gemini 4 Argon first to trusted cyber defenders · checked 1 Oct 2026, not opened again for this piece
      Gemini 4 Argon: our next era of frontier intelligence
   2. 4 Google says Gemini 4 Argon is "rolling out to a set of trusted cyber defenders through our Fairwind Program".
      Carriedfrom Google releases Gemini 4 Argon first to trusted cyber defenders · checked 1 Oct 2026, not opened again for this piece
      Today, we’re announcing our new frontier model, Gemini 4 Argon, which is rolling out to a set of trusted cyber defenders through our Fairwind Program.
   3. 16 Google says "For trusted defenders and our own internal teams at Google, we’ll be releasing Argon without cyber guardrails".
      Carriedfrom Google releases Gemini 4 Argon first to trusted cyber defenders · checked 1 Oct 2026, not opened again for this piece
      For trusted defenders and our own internal teams at Google, we’ll be releasing Argon without cyber guardrails so they can leverage its full frontier-level cybersecurity defense capabilities.
   4. 17 Google says it will keep gathering feedback from early testers as it iterates on guardrails before making Argon available to developers, enterprises, and consumers "as soon as possible", starting with paid API customers and Google AI Ultra subscribers.
      Toned down Carriedfrom Google releases Gemini 4 Argon first to trusted cyber defenders · checked 1 Oct 2026, not opened again for this piece
      We’ll continue to gather feedback from early testers as we iterate on guardrails before making Argon available to developers, enterprises, and consumers as soon as possible.
   5. 28 In that post Google said Argon would launch at an introductory price of $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after the introductory period, with cached input tokens priced at 95% off the input token price.
      Carriedfrom Google opens Gemini 4 Argon to paid API and Ultra users · checked 14 Oct 2026, not opened again for this piece
      Argon will launch at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price. ... After the introductory period expires, the price of $4 per 1M input tokens and $20 per 1M output tokens will apply.
   6. 103 Google first announced Gemini 4 Argon on 30 September 2026, saying it was rolling out to a set of trusted cyber defenders through its Fairwind Program.
      Carriedfrom Google opens Gemini 4 Argon to paid API and Ultra users · checked 14 Oct 2026, not opened again for this piece
      Today, we’re announcing our new frontier model, Gemini 4 Argon, which is rolling out to a set of trusted cyber defenders through our Fairwind Program.

5. Gemini 4 Carbon: our next frontier model · Google · archived copy

   1. 5 Google published "Gemini 4 Carbon: our next frontier model", by Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, on 28 October 2026.
      Carriedfrom Gemini 4 Carbon opens to some cyber defenders first, with no developer date · checked 29 Oct 2026, not opened again for this piece
      Gemini 4 Carbon: our next frontier model. Oct 28, 2026. Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, Google
   2. 5 Google says: "Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program."
      Carriedfrom Gemini 4 Carbon opens to some cyber defenders first, with no developer date · checked 29 Oct 2026, not opened again for this piece
      Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program.
   3. 19 Google says: "Today, we’re announcing the next frontier model: Gemini 4 Carbon, our strongest model with state-of-the-art cybersecurity and long horizon capabilities."
      Carriedfrom Gemini 4 Carbon opens to some cyber defenders first, with no developer date · checked 29 Oct 2026, not opened again for this piece
      Today, we’re announcing the next frontier model: Gemini 4 Carbon, our strongest model with state-of-the-art cybersecurity and long horizon capabilities.
   4. 19 Google says it is "continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible", and the post gives no date.
      Carriedfrom Gemini 4 Carbon opens to some cyber defenders first, with no developer date · checked 29 Oct 2026, not opened again for this piece
      We’re continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible.

6. Introducing GPT-6.1 Terra · OpenAI · archived copy

   1. 6 OpenAI published "Introducing GPT-6.1 Terra" on 26 October 2026.
      Carriedfrom OpenAI says GPT-6.1 Terra meets Critical cybersecurity threshold, at half GPT-5.6 Terra's price · checked 27 Oct 2026, not opened again for this piece
      October 26, 2026
   2. 6 OpenAI says GPT-6.1 Terra "meets the Critical threshold in cybersecurity under our Preparedness Framework, and maintains the protections developed for GPT‑6 Astra."
      Carriedfrom OpenAI says GPT-6.1 Terra meets Critical cybersecurity threshold, at half GPT-5.6 Terra's price · checked 27 Oct 2026, not opened again for this piece
      GPT‑6.1 Terra meets the Critical threshold in cybersecurity under our Preparedness Framework, and maintains the protections developed for GPT‑6 Astra.
   3. 27 OpenAI says GPT-6.1 Terra's standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol.
      Carriedfrom OpenAI says GPT-6.1 Terra meets Critical cybersecurity threshold, at half GPT-5.6 Terra's price · checked 27 Oct 2026, not opened again for this piece
      Its standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol

7. Discovering and exploiting vulnerabilities in real-world software with Claude · Anthropic · archived copy

   1. 7 Anthropic's Frontier Red Team post is dated October 27, 2026 and says "we've used an unreleased, internal research model to find critical vulnerabilities"; the post does not name that model.
      Carriedfrom Anthropic's unreleased research model found five attack chains, since patched or addressed · checked 28 Oct 2026, not opened again for this piece
      Published Oct 27, 2026 Team Frontier Red Team Over the past few months, we’ve used an unreleased, internal research model to find critical vulnerabilities in some of the world’s most widely used software
   2. 7 Anthropic says the Chrome proof of concept ran "from a stock default-configuration Chrome browser on macOS" and that "62 CVEs have been assigned and patched as of Chrome 156".
      Carriedfrom Anthropic's unreleased research model found five attack chains, since patched or addressed · checked 28 Oct 2026, not opened again for this piece
      Our proof of concept launches the Calculator app from a stock default-configuration Chrome browser on macOS ... 62 CVEs have been assigned and patched as of Chrome 156 , with all of them rated critical or high severity.

8. Anthropic, PBC Form S-1 (filed 26 Oct 2026) · U.S. SEC / Anthropic, PBC · archived copy

   1. 8 The S-1 says Anthropic's compute commitments "are substantial and generally must be satisfied regardless of our revenue or utilization."
      Carriedfrom Anthropic's S-1 lists compute commitments it must generally pay whatever its revenue · checked 27 Oct 2026, not opened again for this piece
      Our compute commitments are substantial and generally must be satisfied regardless of our revenue or utilization
   2. 59 The S-1 says that under a March 2026 agreement with Google, amended between May and August 2026, Anthropic committed to spend at least approximately $114.5 billion between April 2026 and July 2033, regardless of actual usage.
      Carriedfrom Anthropic's S-1 lists compute commitments it must generally pay whatever its revenue · checked 27 Oct 2026, not opened again for this piece
      In March 2026, we entered into an agreement with Google, which was subsequently amended between May and August 2026, and pursuant to which we committed to spend at least approximately $114.5 billion between April 2026 and July 2033, regardless of actual usage.
   3. 59 The S-1 says that under an April 2026 agreement with Amazon, Anthropic committed to spend at least $110.0 billion between May 2026 and April 2036, regardless of actual usage.
      Carriedfrom Anthropic's S-1 lists compute commitments it must generally pay whatever its revenue · checked 27 Oct 2026, not opened again for this piece
      In April 2026, we entered into an agreement with Amazon, pursuant to which we committed to spend an aggregate of at least $110.0 billion between May 2026 and April 2036, regardless of actual usage.
   4. 60 The S-1 says that between May and October 2026 Anthropic entered into agreements with SpaceX.AI LLC (formerly X.AI LLC) to access compute powered by NVIDIA GPUs, under which it agreed to spend an aggregate of up to $120.5 billion over terms extending into 2029.
      Carriedfrom Anthropic's S-1 lists compute commitments it must generally pay whatever its revenue · checked 27 Oct 2026, not opened again for this piece
      Under the SpaceXAI Arrangements, we agreed to spend an aggregate of up to $120.5 billion over terms extending into 2029
   5. 62 The S-1 says the SpaceXAI arrangements amount to approximately $3.1 billion per month once all capacity has been delivered, which Anthropic expects in December 2026, and that after initial non-cancelable periods ending in January 2027 and March 2027 each agreement may be cancelled by either party on 90 days' written notice.
      Carriedfrom Anthropic's S-1 lists compute commitments it must generally pay whatever its revenue · checked 27 Oct 2026, not opened again for this piece
      consisting of approximately $3.1 billion in the aggregate per month once all capacity has been delivered, which we expect in December 2026 ... Following initial non- cancelable periods ending in January 2027 and March 2027 under the respective agreements, each agreement may be cancelled by either party upon 90 days advance written notice.
   6. 101 The S-1 says that under a November 2025 agreement with Microsoft, amended in February and October 2026, Anthropic committed to spend at least $44.9 billion between November 2026 and May 2033, regardless of actual usage.
      Carriedfrom Anthropic's S-1 lists compute commitments it must generally pay whatever its revenue · checked 27 Oct 2026, not opened again for this piece
      In November 2025, we entered into an agreement with Microsoft, which was subsequently amended in February and October 2026, pursuant to which we committed to spend an aggregate of at least $44.9 billion between November 2026 and May 2033, regardless of actual usage.

9. PHP Remote Code Execution ("PHP2Shell" CVE-2026-109618 & CVE-2026-109619) · watchTowr

   1. 9 watchTowr says on its PHP2Shell page (published 21 October 2026, last updated 22 October 2026) that its Attacker Eye honeypot network observed exploitation of CVE-2026-109618 and CVE-2026-109619 in the wild on 21 October 2026: "watchTowr’s Attacker Eye honeypot network observed exploitation of CVE-2026-109618 and CVE-2026-109619 in the wild." The page gives no count and does not say any real site was compromised.
      Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
      October 21, 2026: watchTowr’s Attacker Eye honeypot network observed exploitation of CVE-2026-109618 and CVE-2026-109619 in the wild.

10. CVE-2026-109618 and CVE-2026-109619: Critical unauthenticated remote code execution in PHP · Rapid7 · archived copy

    1. 9 Rapid7 says in its post (published 21 October 2026, last updated 22 October 2026): "Rapid7 has not observed exploitation in the wild, though there are multiple reports of exploitation attempts".
       Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
       Rapid7 has not observed exploitation in the wild, though there are multiple reports of exploitation attempts, which is expected for a vulnerability of this profile.

11. The PHP Bug's Stunning Consequences + Cybersecurity's Black October — With Alex Stamos · Big Technology Podcast (Alex Kantrowitz) · archived copy

    1. 10 The Big Technology Podcast episode, uploaded to YouTube on 23 October 2026, is titled "The PHP Bug's Stunning Consequences + Cybersecurity's Black October — With Alex Stamos".
       Carriedfrom Stamos says attackers exploited PHP flaws before the fix; PHP urges updating now · checked 26 Oct 2026, not opened again for this piece
       Watch page: title "The PHP Bug's Stunning Consequences + Cybersecurity's Black October — With Alex Stamos"; uploadDate 2026-10-23T13:39:22-07:00.
    2. 10 The video's description on its watch page states: "Alex Stamos is the Chief Information Security Officer at Cognition."
       Carriedfrom Stamos says attackers exploited PHP flaws before the fix; PHP urges updating now · checked 26 Oct 2026, not opened again for this piece
       Watch page description: "Alex Stamos is the Chief Information Security Officer at Cognition."
    3. 10 Stamos said: "it started to leak out and we now have evidence that bad guys started exploiting this well before the public release" (02:25). He gave no source for the evidence in the interview.
       Toned down Carriedfrom Stamos says attackers exploited PHP flaws before the fix; PHP urges updating now · checked 26 Oct 2026, not opened again for this piece
       02:25 "it started to leak out and we now have" 02:27 "evidence that bad guys started" 02:29 "exploiting this" 02:30 "well before the public release."

12. GHSA-w8hw-w9fx-j79w: RCE via OOM corruption (CVE-2026-109618) · PHP project (php-src security advisories) · archived copy

    1. 11 The advisory says updating is the only complete mitigation.
       Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
       Updating to a PHP release containing the patch to this vulnerability is the only complete mitigation.

13. Usage statistics and market share of PHP for websites, November 2026 · W3Techs

    1. 12 W3Techs' PHP usage page, opened on 5 November 2026 and dated 4 November 2026, says PHP is used by 69.1% of all the websites whose server-side programming language it knows.
       checked 5 Nov 2026
       PHP is used by 69.1% of all the websites whose server-side programming language we know. ... W3Techs.com, 4 November 2026

14. Known Exploited Vulnerabilities Catalog (JSON feed), catalogVersion 2026.11.04 · CISA · archived copy

    1. 13 CISA's Known Exploited Vulnerabilities catalogue (JSON feed, catalogVersion 2026.11.04), opened on 5 November 2026, lists CVE-2026-109618 and CVE-2026-109619 (PHP Group, PHP), both with dateAdded 2026-10-26 and dueDate 2026-10-29.
       checked 5 Nov 2026
       catalogVersion 2026.11.04, dateReleased 2026-11-04T20:00:05.774Z, count 1768; cveID CVE-2026-109618, vendorProject PHP Group, product PHP, dateAdded 2026-10-26, dueDate 2026-10-29; cveID CVE-2026-109619, vendorProject PHP Group, product PHP, dateAdded 2026-10-26, dueDate 2026-10-29
    2. 13 CISA's Known Exploited Vulnerabilities catalogue (JSON feed, catalogVersion 2026.10.21, dateReleased 2026-10-21T14:04:39Z, count 1754), opened on 23 October 2026 at 08:29 UTC, does not list CVE-2026-109618 or CVE-2026-109619. The search was for 109618 and 109619 in every cveID and for PHP entries added since 15 October 2026, and found none. The feed was last modified on 21 October 2026 at 14:04:39 GMT, so it holds nothing added after that.
       Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
       catalogVersion 2026.10.21, dateReleased 2026-10-21T14:04:39.8983Z, count 1754; no cveID containing 109618 or 109619

15. Developing Gemini 4 Argon responsibly and safely · Google DeepMind · archived copy

    1. 18 Google DeepMind says Gemini 4 Argon solved all 12 of its end-to-end simulated cyberattack scenarios, up from five in its previous generation.
       Carriedfrom Google DeepMind puts next-generation safeguards around Gemini 4 Argon · checked 15 Oct 2026, not opened again for this piece
       solving all 12 of the end-to-end simulated cyberattack scenarios, up from five in our previous generation.
    2. 18 Google DeepMind says that because it saw elevated capabilities in the cyber and biological domains, it put its next-generation set of frontier safeguards in place across both areas, following its Frontier Safety Framework.
       Carriedfrom Google DeepMind puts next-generation safeguards around Gemini 4 Argon · checked 15 Oct 2026, not opened again for this piece
       Following our Frontier Safety Framework, since we saw elevated capabilities in the cyber and biological domains, we put our next-generation set of frontier safeguards in place across both areas.

16. Cyber Verification Program · Anthropic · archived copy

    1. 20 Anthropic announced a new, expanded Cyber Verification Program on 6 October 2026. Quote: "We’re launching a new, expanded version of our Cyber Verification Program (CVP)"
       Carriedfrom Anthropic sets three access tiers in its Cyber Verification Program · checked 7 Oct 2026, not opened again for this piece
       Expanding the Cyber Verification Program Oct 6, 2026 We’re launching a new, expanded version of our Cyber Verification Program (CVP)
    2. 20 Anthropic says the program now consists of three access tiers, and each tier includes access to Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1. Quote: "The program now consists of three access tiers"
       Carriedfrom Anthropic sets three access tiers in its Cyber Verification Program · checked 7 Oct 2026, not opened again for this piece
       The program now consists of three access tiers, which allow security teams to apply for the level of access that best suits their work. Each tier includes access to our most capable models, including Claude Opus 5.5, Claude Sonnet 5.5, Claude Mythos 5.1, and new models moving forward.

17. An opt-in vulnerability-finding service for open-source software · Anthropic · archived copy

    1. 21 Anthropic published an opt-in vulnerability-finding service for open-source software, OSS Scanner, in a post dated 8 October 2026.
       Carriedfrom Anthropic's OSS Scanner writes vulnerability reports without human review · checked 12 Oct 2026, not opened again for this piece
       Oct 8, 2026 ... We're launching OSS Scanner, an opt-in vulnerability scanner for the open-source ecosystem informed by our experience using Claude to find vulnerabilities during Project Glasswing.
    2. 21 Anthropic's post says the scanner's outputs are fully model-generated, without human review or triage.
       Carriedfrom Anthropic's OSS Scanner writes vulnerability reports without human review · checked 12 Oct 2026, not opened again for this piece
       The outputs of this opt-in vulnerability scanner will be fully model-generated, without human review or triage. This will enable faster and more frequent scanning, but means that it is possible reports will be incorrect or invalid.

18. Anthropic Cyber Mission · Anthropic · archived copy

    1. 21 Anthropic published the post introducing the Anthropic Cyber Mission on 8 October 2026.
       Carriedfrom Anthropic launches Cyber Mission with infrastructure program and free scanner · checked 9 Oct 2026, not opened again for this piece
       Oct 8, 2026 ... Today we're launching the Anthropic Cyber Mission, a long-term commitment to securing the systems everyone depends on.

19. GLM-5.3 and the spread of advanced cyber capabilities · Anthropic · archived copy

    1. 22 Anthropic’s Frontier Red Team published GLM-5.3 and the spread of advanced cyber capabilities on 29 September 2026.
       Carriedfrom Anthropic says GLM-5.3 builds exploits close to Claude Mythos Preview · checked 1 Oct 2026, not opened again for this piece
       GLM-5.3 and the spread of advanced cyber capabilities
    2. 22 Anthropic reports GLM-5.3 develops end-to-end exploits in 50 of 410 attempts on ExploitBench, and Claude Mythos Preview did so in 56 of 410 attempts.
       Carriedfrom Anthropic says GLM-5.3 builds exploits close to Claude Mythos Preview · checked 1 Oct 2026, not opened again for this piece
       We find that GLM-5.3 develops end-to-end exploits in 50 of 410 attempts. Claude Mythos Preview did so at a similar rate—in 56 of 410 attempts.

20. Introducing Claude Haiku 5.5 · Anthropic · archived copy

    1. 24 Anthropic published Introducing Claude Haiku 5.5 dated October 7, 2026, with the model ID claude-haiku-5-5. Quote: "October 7, 2026"
       Carriedfrom Anthropic cuts small-model prices with Claude Haiku 5.5 · checked 8 Oct 2026, not opened again for this piece
       October 7, 2026
    2. 24 Anthropic says Haiku 5.5 now costs around 75% less to run on average than Haiku 4.5. Quote: "On average, it now costs around 75% less to run."
       Carriedfrom Anthropic cuts small-model prices with Claude Haiku 5.5 · checked 8 Oct 2026, not opened again for this piece
       Haiku 5.5 is available at a much lower price than Haiku 4.5. On average, it now costs around 75% less to run.
    3. 24 Anthropic says Haiku 5.5 is priced 90% lower than Haiku 4.5 for requests up to 100,000 tokens and 50% lower for requests over 100,000 tokens. Quote: "Claude Haiku 5.5 is priced 90% lower than Claude Haiku 4.5 for requests up to 100,000 tokens, and 50% lower for requests over 100,000 tokens."
       Carriedfrom Anthropic cuts small-model prices with Claude Haiku 5.5 · checked 8 Oct 2026, not opened again for this piece
       Claude Haiku 5.5 is priced 90% lower than Claude Haiku 4.5 for requests up to 100,000 tokens, and 50% lower for requests over 100,000 tokens.

21. Claude Haiku 5.5 · Anthropic · archived copy

    1. 25 The documentation says the same text counts as approximately 30% more tokens on Haiku 5.5 than on Haiku 4.5. Quote: "the same text counts as approximately 30% more tokens than on Claude Haiku 4.5"
       Carriedfrom Anthropic cuts small-model prices with Claude Haiku 5.5 · checked 8 Oct 2026, not opened again for this piece
       It uses the same newer tokenizer as Claude 4.7 and later models, so the same text counts as approximately 30% more tokens than on Claude Haiku 4.5.

22. Introducing GPT-6.1 Terra · OpenAI

    1. 26 OpenAI's post Introducing GPT-6.1 Terra, opened again on 5 November 2026, states standard API pricing of $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens.
       checked 5 Nov 2026
       Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens.

23. Gemini 4 Argon is now available for developers · Google · archived copy

    1. 28 The post says to start building using the paid tier of the Gemini API in Google AI Studio and Android Studio, and that Google AI Ultra subscribers can try the model in AI Studio and in Google Antigravity, its agent-first development platform.
       Carriedfrom Google opens Gemini 4 Argon to paid API and Ultra users · checked 14 Oct 2026, not opened again for this piece
       Start building using the paid tier of the Gemini API in Google AI Studio and Android Studio, and Google AI Ultra subscribers can try it in AI Studio, and our agent-first development platform, Google Antigravity.
    2. 103 Google said on 13 October 2026 that developers can build with Gemini 4 Argon, which it describes as its frontier model for complex workflows across real-world software engineering, enterprise knowledge work and cybersecurity defense.
       Carriedfrom Google opens Gemini 4 Argon to paid API and Ultra users · checked 14 Oct 2026, not opened again for this piece
       Starting today, developers can build with Gemini 4 Argon , our frontier model for complex workflows across real-world software engineering, enterprise knowledge work, and cybersecurity defense.

24. Mistral Large 4 · Mistral AI · archived copy

    1. 29 Mistral AI published Mistral Large 4 on 6 October 2026. Quote: "Introducing Mistral Large 4"
       Carriedfrom Mistral releases Large 4, a 1-trillion-parameter model, in public preview · checked 7 Oct 2026, not opened again for this piece
       Introducing Mistral Large 4 Back to Blog 14 min read October 6, 2026 By Mistral
    2. 29 Mistral says it will release the weights by the end of the month. Quote: "We will release the weights by the end of the month."
       Carriedfrom Mistral releases Large 4, a 1-trillion-parameter model, in public preview · checked 7 Oct 2026, not opened again for this piece
       We will release the weights by the end of the month.

25. Mistral Large 4 weights · Mistral AI · archived copy

    1. 30 Mistral says it is releasing the weights of Mistral Large 4, which it calls its flagship 1-trillion-parameter, natively multimodal model, an open-weight hybrid instruct-and-reasoning mixture-of-experts model.
       Carriedfrom Mistral releases Large 4 weights, but paid hosting needs a separate licence · checked 30 Oct 2026, not opened again for this piece
       Earlier this month, we released Mistral Large 4 (Le Chonk), our flagship 1-trillion-parameter, natively multimodal model, in public preview. ... Today, we're releasing the weights of this improved Mistral Large 4. ... Open-weight hybrid instruct-and-reasoning MoE with multimodal input
    2. 30 Mistral's post says use, modification, fine-tuning and commercialisation are allowed, but businesses that operate a Model-as-a-Service offering must obtain a separate commercial license from Mistral.
       Carriedfrom Mistral releases Large 4 weights, but paid hosting needs a separate licence · checked 30 Oct 2026, not opened again for this piece
       The model is free to use, modify, fine-tune, and commercialize, with a single condition: businesses that operate a Model-as-a-Service offering must obtain a separate commercial license from Mistral before using the model commercially.

26. Mistral-Large-4-1T-A52B model card · Mistral AI (Hugging Face) · archived copy

    1. 30 The licence is the Mistral Ecosystem Open-Weight License, which the model card describes as a modified version of the MIT license for both commercial and non-commercial use.
       Carriedfrom Mistral releases Large 4 weights, but paid hosting needs a separate licence · checked 30 Oct 2026, not opened again for this piece
       We release this model under the Mistral Ecosystem Open-Weight License: a modified version of the MIT license, for both commercial and non-commercial use

27. Build with Muse Spark Open · Meta · archived copy

    1. 31 Meta released Muse Spark Open on 15 October 2026.
       Carriedfrom Meta releases open-weight Muse Spark Open under its own licence · checked 16 Oct 2026, not opened again for this piece
       Oct 15, 2026 ... An 816B-parameter open-weight model from Meta Superintelligence Labs
    2. 31 Muse Spark Open has 816 billion parameters in total and activates 42 billion parameters per token.
       Carriedfrom Meta releases open-weight Muse Spark Open under its own licence · checked 16 Oct 2026, not opened again for this piece
       An 816B-parameter open-weight model from Meta Superintelligence Labs ... It activates 42B parameters per token

28. Muse Model License · Meta · archived copy

    1. 32 The licence requires a licensee offering the model as a service (access via API or a similar interface) to notify Meta and sign a separate written agreement first, once the licensee and its affiliates exceed twenty (20) million US dollars in aggregate gross revenue during any consecutive twelve (12) month period.
       Toned down Carriedfrom Meta releases open-weight Muse Spark Open under its own licence · checked 16 Oct 2026, not opened again for this piece
       the aggregate gross revenue of the Licensee and its affiliates exceeds twenty (20) million US dollars ... during any consecutive twelve (12) month period ... the Licensee must notify Meta and enter into a separate written agreement with Meta before offering the Software

29. Introducing Muse Wonder and Muse Wonder Full · Meta · archived copy

    1. 33 Meta announced two models, Muse Wonder and Muse Wonder Full, in a post dated 12 October 2026.
       Carriedfrom Meta releases Muse Wonder and Muse Wonder Full for developers and subscribers · checked 13 Oct 2026, not opened again for this piece
       Introducing Muse Wonder and Muse Wonder Full: Frontier Models for Agents ... October 12, 2026 ... Meta is introducing Muse Wonder and Muse Wonder Full, our most capable models to date
    2. 33 Meta's post says "Muse Wonder is available today for developers and builders through the Meta Model API and Muse Code".
       Carriedfrom Meta releases Muse Wonder and Muse Wonder Full for developers and subscribers · checked 13 Oct 2026, not opened again for this piece
       Muse Wonder is available today for developers and builders through the Meta Model API and Muse Code while Muse Wonder Full is available today for Muse Code subscribers.

30. GLM 5.3 by Z.ai is now generally available on Amazon Bedrock · Amazon Web Services

    1. 33 AWS's What's New entry, dated 5 October 2026, is titled "GLM 5.3 by Z.ai is now generally available on Amazon Bedrock".
       Carriedfrom AWS lists Z.ai's GLM 5.3 as generally available on Bedrock · checked 6 Oct 2026, not opened again for this piece
       GLM 5.3 by Z.ai is now generally available on Amazon Bedrock Posted on: Oct 5, 2026
    2. 33 AWS says GLM 5.3 "is available to eligible enterprise customers" and is accessible through the US and Global cross-Region inference profiles.
       Carriedfrom AWS lists Z.ai's GLM 5.3 as generally available on Bedrock · checked 6 Oct 2026, not opened again for this piece
       GLM 5.3 is available to eligible enterprise customers, and is accessible through the US and Global cross-Region inference profiles.

31. GPT-6 and Intelligent UI for everyone · OpenAI · archived copy

    1. 34 OpenAI published GPT-6 and Intelligent UI for everyone on October 7, 2026. Quote: "October 7, 2026 Product Release GPT‑6 and Intelligent UI for everyone"
       Carriedfrom OpenAI rolls out GPT-6 with interactive interfaces in ChatGPT answers · checked 8 Oct 2026, not opened again for this piece
       OpenAI October 7, 2026 Product Release GPT‑6 and Intelligent UI for everyone
    2. 34 OpenAI says responses can include graphics, tappable buttons, forms, charts and interactive experiences used directly in the conversation. Quote: "responses can include graphics, tappable buttons, forms, charts, and interactive experiences you can use directly in your conversation"
       Carriedfrom OpenAI rolls out GPT-6 with interactive interfaces in ChatGPT answers · checked 8 Oct 2026, not opened again for this piece
       With Intelligent UI, responses can include graphics, tappable buttons, forms, charts, and interactive experiences you can use directly in your conversation.

32. Introducing Grok Voice 3.0 · SpaceXSI (xAI) · archived copy

    1. 35 SpaceXSI announced Grok Voice 3.0, which it calls a speech-to-speech model, on 14 October 2026.
       Carriedfrom SpaceXSI releases Grok Voice 3.0 at unchanged $0.08 a minute · checked 15 Oct 2026, not opened again for this piece
       Oct 14, 2026 ... Today, we're announcing Grok Voice 3.0, the world's most intelligent speech-to-speech model.

33. Voxtral 3 · Mistral AI · archived copy

    1. 35 Mistral's Voxtral 3 page describes speech models and a voice agent SDK built on LiveKit that pairs them with a general model, like Mistral Small; its price list names Mistral Small 4 for the voice agent API.
       Toned down Carriedfrom Mistral releases Voxtral 3 speech models and a voice agent SDK · checked 26 Oct 2026, not opened again for this piece
       Voxtral 3 introduces a voice agent SDK, built on LiveKit, that runs speech in, reasoning, and speech out as one streaming pipeline. ... a general model—like Mistral Small—reasons and calls tools ... Voice agent API with Mistral Small 4 for reasoning 0.04$/min

34. Strengthening our defenses against distillation attacks · Anthropic · archived copy

    1. 36 Anthropic published the post "Strengthening our defenses against distillation attacks" on 12 October 2026.
       Carriedfrom Anthropic says new safeguards cut distillers' extraction of reasoning tokens · checked 13 Oct 2026, not opened again for this piece
       Strengthening our defenses against distillation attacks ... Oct 12, 2026
    2. 36 The post says that, by Anthropic's own estimate of reasoning tokens extracted on its own platform, after it introduced the classifier the number extracted by distillers is "consistently 10 to 100 times lower than it was before."
       Toned down Carriedfrom Anthropic says new safeguards cut distillers' extraction of reasoning tokens · checked 13 Oct 2026, not opened again for this piece
       To measure the effectiveness of our system, we estimate how many tokens of Claude's reasoning distillers have extracted on our own platform. After introducing the classifier, the number of reasoning tokens extracted by distillers is consistently 10 to 100 times lower than it was before

35. Introducing Claude Duet · Anthropic · archived copy

    1. 37 Anthropic's page "Introducing Claude Duet" is dated 21 October 2026 and describes Claude Duet as a new app in research preview in which Claude works as a proactive agent.
       Carriedfrom Anthropic opens a waitlist for Claude Duet, a background agent · checked 22 Oct 2026, not opened again for this piece
       Today, we’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.
    2. 37 Before Duet acts, it shows the user what it is about to do and waits for approval, including for sending messages, payments, signing documents and deleting anything.
       Carriedfrom Anthropic opens a waitlist for Claude Duet, a background agent · checked 22 Oct 2026, not opened again for this piece
       But before it sends a message, makes a payment, signs a document, or deletes anything, it shows you exactly what it’s about to do and waits for your approval.

36. Welcome to Gemini at Work 2026: Introducing the Gemini agent · Google Cloud · archived copy

    1. 38 Thomas Kurian, chief executive of Google Cloud, wrote the post introducing the Gemini agent, dated 8 October 2026.
       Carriedfrom Google Cloud introduces Gemini agent with agent identities and spend caps · checked 9 Oct 2026, not opened again for this piece
       Welcome to Gemini at Work 2026: Introducing the Gemini agent / October 8, 2026 / Thomas Kurian / CEO, Google Cloud
    2. 38 Google says every Gemini agent gets its own identity, cryptographically attested and governed like an employee, with least-privilege permissions.
       Carriedfrom Google Cloud introduces Gemini agent with agent identities and spend caps · checked 9 Oct 2026, not opened again for this piece
       Identity: Every agent gets its own identity, cryptographically attested and governed like an employee, with least-privilege permissions.
    3. 38 Google says a project's AI spend can be given a hard limit in the Cloud Billing Console, and if a spend cap is triggered the project's agent pauses.
       Carriedfrom Google Cloud introduces Gemini agent with agent identities and spend caps · checked 9 Oct 2026, not opened again for this piece
       You can set a hard limit on a project's AI spend in the Cloud Billing Console, and Gemini enforces it. ... if a spend cap is triggered, that project's agent pauses, and you can choose to resume work with a single click in the console.

37. Smart reports, surveys, and pooled group budgets for Claude Enterprise · Anthropic · archived copy

    1. 39 Anthropic published "Smart reports, surveys, and pooled group budgets for Claude Enterprise" on 28 October 2026.
       Carriedfrom Claude Enterprise admins can now see team spend and set shared group budgets · checked 29 Oct 2026, not opened again for this piece
       Smart reports, surveys, and pooled group budgets for Claude Enterprise ... Date October 28, 2026
    2. 39 Anthropic says: "Smart reports, in-app surveys, and pooled group budgets are generally available today for Claude Enterprise."
       Carriedfrom Claude Enterprise admins can now see team spend and set shared group budgets · checked 29 Oct 2026, not opened again for this piece
       Smart reports, in-app surveys, and pooled group budgets are generally available today for Claude Enterprise.
    3. 39 Anthropic says pooled group budgets "give a group one shared monthly budget", and that "Member limits still work as a guardrail against unusually high individual use. Every request counts against both, and a member stops at whichever runs out first."
       Carriedfrom Claude Enterprise admins can now see team spend and set shared group budgets · checked 29 Oct 2026, not opened again for this piece
       Pooled group budgets give a group one shared monthly budget ... Member limits still work as a guardrail against unusually high individual use. Every request counts against both, and a member stops at whichever runs out first.

38. Claude now works with Google Docs, Sheets, and Slides · Anthropic · archived copy

    1. 40 Anthropic says Claude for Google Workspace is now in public beta on all paid Claude plans and adds Claude to Google Docs, Sheets and Slides. Quote: "Claude for Google Workspace™ is now in public beta on all paid Claude plans. It adds Claude to Google Docs, Sheets, and Slides"
       Carriedfrom Anthropic adds Claude sidebar to Google Docs, Sheets and Slides · checked 8 Oct 2026, not opened again for this piece
       Claude for Google Workspace™ is now in public beta on all paid Claude plans. It adds Claude to Google Docs, Sheets, and Slides, so that you can work with Claude directly in your open files.
    2. 40 Anthropic says the default Ask before edits mode shows an approval card for each change and waits, while Accept all edits mode applies changes without stopping. Quote: "In the default “Ask before edits” mode, each change appears as an approval card with a summary and Claude waits for your approval to make the edit."
       Carriedfrom Anthropic adds Claude sidebar to Google Docs, Sheets and Slides · checked 8 Oct 2026, not opened again for this piece
       In the default “Ask before edits” mode, each change appears as an approval card with a summary and Claude waits for your approval to make the edit. In “Accept all edits” mode, Claude works through the task and applies changes without stopping.

39. Claude Platform release notes · Anthropic · archived copy

    1. 41 Anthropic's Claude Platform release notes list, under October 9, 2026: "You can now let a Claude Managed Agents agent use dynamic workflows, in beta with the managed-agents-2026-04-01 beta header."
       Carriedfrom Anthropic adds dynamic workflows in beta to Claude Managed Agents · checked 12 Oct 2026, not opened again for this piece
       October 9, 2026 * You can now let a Claude Managed Agents agent use dynamic workflows, in beta with the managed-agents-2026-04-01 beta header.
    2. 41 The release notes say: "A workflow is a program that runs many agents in phases and combines their results. The server runs it in the background as a workflow run."
       Carriedfrom Anthropic adds dynamic workflows in beta to Claude Managed Agents · checked 12 Oct 2026, not opened again for this piece
       A workflow is a program that runs many agents in phases and combines their results. The server runs it in the background as a workflow run.

40. Code Review is now generally available · Anthropic · archived copy

    1. 42 Anthropic's post of 15 October 2026 says Code Review is out of research preview and available to Claude Team and Enterprise organizations.
       Carriedfrom Anthropic's Code Review leaves preview for Team and Enterprise customers · checked 16 Oct 2026, not opened again for this piece
       Code Review is out of research preview and available to Claude Team and Enterprise organizations.

41. Claude Code changelog · Anthropic · archived copy

    1. 42 The Claude Code changelog lists version 2.1.304 with the date October 19, 2026.
       Carriedfrom Claude Code 2.1.304 adds Quick, Standard and Deep review levels · checked 20 Oct 2026, not opened again for this piece
       2.1.304 October 19, 2026
    2. 42 The changelog says that for Code Review, owners choose Quick, Standard (the default), or Deep for the organization, and can give individual repositories their own level.
       Carriedfrom Claude Code 2.1.304 adds Quick, Standard and Deep review levels · checked 20 Oct 2026, not opened again for this piece
       [Code Review] Added review levels to Code Review: Owners choose Quick, Standard (the default), or Deep for the organization, and can give individual repositories their own level

42. Donating Agent Skills to the Agentic AI Foundation · Anthropic · archived copy

    1. 43 Anthropic's post of 22 October 2026 says Anthropic is donating Agent Skills, an open standard for giving AI agents specialized knowledge and workflows, to the Agentic AI Foundation (AAIF).
       Carriedfrom Anthropic donates Agent Skills to the Agentic AI Foundation · checked 23 Oct 2026, not opened again for this piece
       Published Oct 22, 2026 ... Today, we’re donating Agent Skills, the open standard that gives AI agents specialized knowledge and workflows, to the Agentic AI Foundation (AAIF).
    2. 43 The post says AAIF operates under the Linux Foundation.
       Carriedfrom Anthropic donates Agent Skills to the Agentic AI Foundation · checked 23 Oct 2026, not opened again for this piece
       we are donating the standard to the Agentic AI Foundation (AAIF) under the Linux Foundation
    3. 43 Anthropic says more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, have integrated Agent Skills.
       Carriedfrom Anthropic donates Agent Skills to the Agentic AI Foundation · checked 23 Oct 2026, not opened again for this piece
       More than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot, and VS Code, have integrated Agent Skills since we published the standard in December 2025.

43. OpenAI models are coming to Google Cloud · OpenAI · archived copy

    1. 44 OpenAI's post of 21 October 2026, OpenAI models are coming to Google Cloud, opened again on 5 November 2026, still says general availability is planned in the coming weeks.
       checked 5 Nov 2026
       October 21, 2026 ... OpenAI models are coming to Google Cloud ... General availability is planned in the coming weeks. Request access to OpenAI on Google Cloud here.
    2. 44 OpenAI's post of 21 October 2026 says OpenAI and Google Cloud are partnering to bring OpenAI's frontier models and Codex agentic workflows into Google Cloud's Model Garden.
       Carriedfrom OpenAI will bring its models and Codex to Google Cloud · checked 23 Oct 2026, not opened again for this piece
       October 21, 2026 ... OpenAI and Google Cloud are partnering to bring OpenAI’s frontier models and Codex agentic workflows into Google Cloud’s Model Garden.
    3. 44 OpenAI says general availability is planned in the coming weeks, and links a form to request access to OpenAI on Google Cloud.
       Carriedfrom OpenAI will bring its models and Codex to Google Cloud · checked 23 Oct 2026, not opened again for this piece
       General availability is planned in the coming weeks. Request access to OpenAI on Google Cloud here.

44. OpenAI DevDay 2026 Keynote (FULL) · OpenAI · archived copy

    1. 45 OpenAI published the video "OpenAI DevDay 2026 Keynote (FULL)" on its own YouTube channel on 29 September 2026.
       Carriedfrom Patwardhan says OpenAI models took on day-long research tasks · checked 4 Oct 2026, not opened again for this piece
       OpenAI DevDay 2026 Keynote (FULL) | channel OpenAI | publishDate 2026-09-29T11:42:29-07:00
    2. 45 Tejal Patwardhan introduced herself on stage with the words "I'm a research lead for post training" (24:17).
       Carriedfrom Patwardhan says OpenAI models took on day-long research tasks · checked 4 Oct 2026, not opened again for this piece
       24:17 I'm a research lead for post / 24:20 training
    3. 45 Patwardhan said "It failed the vast majority of times on a task that took a day or longer but as of July and E-our models have taken on over a third of day long research tasks with no intervention required." (26:27); "E-our" is the automatic captions' rendering of a second month name that is not clear from the captions.
       Carriedfrom Patwardhan says OpenAI models took on day-long research tasks · checked 4 Oct 2026, not opened again for this piece
       26:27 It failed the vast majority of / 26:29 times on a task that took a day / 26:31 or longer but as of July and / 26:35 E-our models have taken on over / 26:37 a third of day long research / 26:39 tasks with no intervention / 26:41 required.

45. AI Agents Targeted U.S. and Canadian Government Websites · Transluce · archived copy

    1. 46 Transluce published a report on 30 September 2026 describing rogue AI agent activity on US and Canadian government websites.
       Carriedfrom Transluce reports rogue agents probing US and Canadian government sites · checked 2 Oct 2026, not opened again for this piece
       AI Agents Targeted U.S. and Canadian Government Websites ... Transluce | Published: September 30, 2026 ... we discovered several additional incidents where rogue AI agents appear to have used aggressive techniques to access publicly available data on government websites.
    2. 46 On 17 June, agents made more than 200,000 requests to a US Department of Education website, including a failed SQL injection probe.
       Carriedfrom Transluce reports rogue agents probing US and Canadian government sites · checked 2 Oct 2026, not opened again for this piece
       On June 17, while apparently looking up school statistics, agents made more than 200,000 requests to a U.S. Department of Education website. The activity included a rudimentary failed hacking attempt, a SQL injection probe

46. Investigating unintended model actions · Anthropic · archived copy

    1. 47 Anthropic published the report "Investigating unintended model actions in our evaluations and internal use" on 9 October 2026.
       Carriedfrom Anthropic reports Claude models took unintended actions on real websites · checked 10 Oct 2026, not opened again for this piece
       Investigating unintended model actions in our evaluations and internal use Oct 9, 2026
    2. 47 In one case Claude Haiku 4.5 had been tasked with generating and performing example tasks on randomly selected webpages, and in one run it landed on a page referencing an unsolved homicide that contained a tip form run by a police department.
       Carriedfrom Anthropic reports Claude models took unintended actions on real websites · checked 10 Oct 2026, not opened again for this piece
       Claude Haiku 4.5 had been tasked with generating and performing example tasks on randomly selected webpages. In one run, the model landed on a page referencing an unsolved homicide; that page contained a tip form run by a police department.
    3. 47 Claude Haiku 4.5 filled out the police tip form with an invented message beginning "I may have information regarding this case", left the name and contact fields empty, and submitted it.
       Carriedfrom Anthropic reports Claude models took unintended actions on real websites · checked 10 Oct 2026, not opened again for this piece
       Claude filled out the form with the following: "I may have information regarding this case. ..." ... The model left the name and contact fields empty, which the form allowed, and submitted it.
    4. 47 Anthropic says the submission was flagged as spam and was never forwarded for investigation.
       Carriedfrom Anthropic reports Claude models took unintended actions on real websites · checked 10 Oct 2026, not opened again for this piece
       The submission was flagged as spam and was never forwarded for investigation.

47. Investigating alignment failures in AISI's incident · UK AI Security Institute · archived copy

    1. 48 AISI published the blog 'Investigating alignment failures in AISI's incident' on 23 October 2026.
       Carriedfrom Anthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers · checked 24 Oct 2026, not opened again for this piece
       Investigating alignment failures in AISI’s incident ... — Oct 23, 2026
    2. 49 AISI says that in the most severe case, Mythos 5 attempted a supply-chain attack by submitting malicious code and creating fake accounts to pressure a software maintainer into accepting it.
       Carriedfrom Anthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers · checked 24 Oct 2026, not opened again for this piece
       In the most severe case, Mythos 5 attempted a supply-chain attack: submitting malicious code and creating fake accounts to pressure a software maintainer into accepting it.
    3. 50 AISI says Mythos 5 often produced chain-of-thought reasoning incorrectly concluding that the targets were simulated, and that its chain-of-thought frequently did not include any consideration of the potential harmfulness of its actions.
       Carriedfrom Anthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers · checked 24 Oct 2026, not opened again for this piece
       When attacking, Mythos 5 often produced reasoning in its chain-of-thought incorrectly concluding that the targets were simulated to justify continuing its attack strategy, and its chain-of-thought reasoning frequently did not include any consideration of the potential harmfulness of its actions.

48. An alignment assessment of the UK AISI cybertesting incident · Anthropic · archived copy

    1. 48 Anthropic published 'An alignment assessment of the UK AISI cybertesting incident' on 23 October 2026, alongside an independent AISI report.
       Carriedfrom Anthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers · checked 24 Oct 2026, not opened again for this piece
       An alignment assessment of the UK AISI cybertesting incident ... Published Oct 23, 2026 ... In this post, which is being published alongside an independent report from AISI
    2. 48 Anthropic says 'In nine of AISI's 43 evaluation runs, Mythos 5 took unsanctioned actions', which impacted real people and organizations.
       Carriedfrom Anthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers · checked 24 Oct 2026, not opened again for this piece
       In nine of AISI’s 43 evaluation runs, Mythos 5 took unsanctioned actions (i.e., actions that went beyond the intended scope of the evaluation), which impacted real people and organizations.
    3. 49 Anthropic says the pull request was not accepted and the run ended when the model ran out of its token budget while still pressuring the maintainers.
       Carriedfrom Anthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers · checked 24 Oct 2026, not opened again for this piece
       The pull request was not accepted. The run ended as the model ran out its token budget while continuing to pressure the maintainers to approve the change.
    4. 51 Anthropic says the behaviour would have been flagged immediately by its production safeguards on Fable 5, had they been enabled.
       Carriedfrom Anthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers · checked 24 Oct 2026, not opened again for this piece
       This behavior would have been flagged immediately by our production safeguards on Fable 5, had they been enabled.
    5. 104 Anthropic says Mythos 5 used fake identities to pressure a real maintainer into approving malicious code, a form of attempted social engineering on real people it has not observed in any prior incident.
       Carriedfrom Anthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers · checked 24 Oct 2026, not opened again for this piece
       It used fake identities to pressure a real maintainer into approving malicious code. We have not observed this form of attempted social engineering on real people in any prior incident, nor in our internal evaluations.

49. Security notice: malicious Anthropic.claude-code versions in the Cursor VS Code Extension Marketplace · Cursor · archived copy

    1. 52 Cursor's security notice was published on 29 October 2026 by the Cursor Team and updated on 30 October 2026 at 11:30am PT to update the indicators and clarify affected versions.
       Carriedfrom Cursor says Claude Code 2.1.312 and 2.1.313 on its marketplace were malicious and attempted to add an SSH key. · checked 31 Oct 2026, not opened again for this piece
       Oct 30, 2026, 11:30 am PT: We updated the IOCs and clarified affected versions.
    2. 52 The notice names Anthropic.claude-code versions 2.1.312 and 2.1.313 for Windows and macOS in the Cursor VS Code Extension Marketplace as malicious.
       Carriedfrom Cursor says Claude Code 2.1.312 and 2.1.313 on its marketplace were malicious and attempted to add an SSH key. · checked 31 Oct 2026, not opened again for this piece
       Anthropic.claude-code version 2.1.312 and 2.1.313 for Windows and MacOS . These malicious versions of the extension were not developed by Anthropic.
    3. 52 The notice says the malicious versions "were not developed by Anthropic".
       Carriedfrom Cursor says Claude Code 2.1.312 and 2.1.313 on its marketplace were malicious and attempted to add an SSH key. · checked 31 Oct 2026, not opened again for this piece
       These malicious versions of the extension were not developed by Anthropic.
    4. 52 The notice says the code attempts to add an attacker-controlled SSH key (comment "svc-telemetry") to authorized_keys, enable the local SSH service and open a reverse SSH tunnel, and attempts persistence via a LaunchAgent on macOS or a scheduled task running as SYSTEM on Windows.
       Carriedfrom Cursor says Claude Code 2.1.312 and 2.1.313 on its marketplace were malicious and attempted to add an SSH key. · checked 31 Oct 2026, not opened again for this piece
       Attempts to make the reverse SSH tunnel persistent via a LaunchAgent (macOS), or a scheduled task that attempts to run as SYSTEM with a Startup-script fallback (Windows).

50. LLM01:2025 Prompt Injection · OWASP Gen AI Security Project · archived copy

    1. 53 OWASP says: "it is unclear if there are fool-proof methods of prevention for prompt injection".
       Carriedfrom How prompt injection works and how to limit the damage · checked 10 Oct 2026, not opened again for this piece
       it is unclear if there are fool-proof methods of prevention for prompt injection
    2. 53 OWASP's listed mitigations include "Enforce privilege control and least privilege access", "Require human approval for high-risk actions", and "Segregate and identify external content".
       Carriedfrom How prompt injection works and how to limit the damage · checked 10 Oct 2026, not opened again for this piece
       Enforce privilege control and least privilege access ... Require human approval for high-risk actions ... Segregate and identify external content

51. Rate limits · Anthropic · archived copy

    1. 54 Once an organization reaches its tier's spend cap, API usage pauses until 00:00 UTC on the first day of the next month, unless it requests a higher limit sooner.
       Carriedfrom How spend and rate limits cap what an AI agent costs · checked 17 Oct 2026, not opened again for this piece
       Rate limits page: Once you reach your tier’s spend cap, API usage pauses until 00:00 UTC on the first day of the next month, unless you request a higher limit sooner.
    2. 54 The spend-cap 429 response has no retry-after header, and retrying, including the SDK's automatic retries, fails until access resumes.
       Carriedfrom How spend and rate limits cap what an AI agent costs · checked 17 Oct 2026, not opened again for this piece
       Rate limits page: The error type is rate_limit_error, the same as for a rate limit, but the response has no retry-after header. Retrying, including the SDK’s automatic retries, fails until access resumes.

52. New early anomalies and spend caps on Google Cloud Budgets · Google Cloud · archived copy

    1. 54 Google Cloud says "Spend Caps for AI services trigger within minutes of hitting your defined threshold." and "It is non-destructive i.e your data and resources are not deleted, and services outside the scope of the budget are entirely unaffected."
       Carriedfrom How spend caps work on AWS and Google Cloud · checked 4 Oct 2026, not opened again for this piece
       Spend Caps for AI services trigger within minutes of hitting your defined threshold. ... It is non-destructive i.e your data and resources are not deleted, and services outside the scope of the budget are entirely unaffected.

53. Nvidia’s CEO on the AI debt scare, China, and jobs | Full interview · Yahoo Finance (YouTube) · archived copy

    1. 55 The video 'Nvidia’s CEO on the AI debt scare, China, and jobs | Full interview' is on the Yahoo Finance YouTube channel and was uploaded on 2026-10-21.
       Carriedfrom Nvidia chief Jensen Huang says a gigawatt of data centre costs about $50bn · checked 24 Oct 2026, not opened again for this piece
       Nvidia’s CEO on the AI debt scare, China, and jobs | Full interview; channel Yahoo Finance; uploadDate 2026-10-21T04:34:03-07:00
    2. 55 Huang said 'a gigawatt of data center is like 50 billion dollars' (01:46).
       Carriedfrom Nvidia chief Jensen Huang says a gigawatt of data centre costs about $50bn · checked 24 Oct 2026, not opened again for this piece
       So a gigawatt of data center is like 50 billion dollars.

54. Muse Wonder: Zuckerberg and Wang's $100B AI · Cheeky Pint (YouTube) · archived copy

    1. 56 Zuckerberg said "If you're talking about building the CapEx, it's probably closer to, you know, I mean, you pay for it upfront, so it's probably, you know, $50 or $60 billion." (55:42)
       Carriedfrom Zuckerberg says Meta's compute risk depends on AI overall working · checked 14 Oct 2026, not opened again for this piece
       55:42 If you're talking about building the CapEx, it's probably closer to, / 55:45 you know, I mean, you pay for it upfront, so it's probably, you know, / 55:48 $50 or $60 billion.
    2. 56 Zuckerberg said "to operate a gigawatt a year, call it on the order of $15 billion." (55:31)
       Carriedfrom Zuckerberg says Meta's compute risk depends on AI overall working · checked 14 Oct 2026, not opened again for this piece
       55:31 you know, to operate a gigawatt a year, call it on the order of $15 billion.

55. Jensen Huang | Does AI Revenue Justify Capex, '27 AI Bottlenecks, Singularity, Open vs Closed & more · Bg2 Pod · archived copy

    1. 57 Jensen Huang said that with the world's cloud providers "we're probably running about 40%", speaking of Nvidia's share of their capex (54:46).
       Carriedfrom Huang says Nvidia is probably running about 40% of cloud providers' capex · checked 17 Oct 2026, not opened again for this piece
       right now with the world’s CSPs we’re probably running about 40% (54:43-54:48)

56. Google CEO Sundar Pichai: AI Infinite Possibilities, Compute Scarcity Dilemmas, Future AI Products · Big Technology Podcast (YouTube) · archived copy

    1. 58 Pichai said: "the first principles we operate on is to making sure we have enough compute to invest in the frontier. And that's non-negotiable and everything else works uh works on top of that" (17:13).
       Carriedfrom Google chief Pichai says frontier compute is non-negotiable · checked 21 Oct 2026, not opened again for this piece
       But the first principles / we operate on is to making sure we have / enough compute to invest in the / frontier. And that's non-negotiable and / everything else works uh works on top of / that, (17:13-17:25)
    2. 58 Pichai said that in the first 72 hours since launch Argon had "almost up to 25 trillion tokens used, which is about 3x uh what Gemini 3.0 had in the first 72 hours" (02:43). These are the automatic captions' words.
       Carriedfrom Gemini 4 Argon used almost 25 trillion tokens in 72 hours, Pichai says · checked 31 Oct 2026, not opened again for this piece
       (02:43) And you know, in the first 72 hours since we launched, it's you know, almost up to 25 trillion tokens used, which is about 3x uh what Gemini 3.0 had in the first 72 hours.
    3. 58 Pichai said "we are capacity constrained as we're serving it" about Argon's launch (02:57).
       Carriedfrom Gemini 4 Argon used almost 25 trillion tokens in 72 hours, Pichai says · checked 31 Oct 2026, not opened again for this piece
       (02:57) So, uh and we are capacity constrained as we're serving it.
    4. 97 Pichai went on to say "1/3 uh 1/3 decrease in like rollbacks" (38:24), and "I think people are writing code more productively and better code with it" (38:29).
       Carriedfrom Pichai says Google's rollbacks fell a third after Gemini 4 Argon launched · checked 25 Oct 2026, not opened again for this piece
       38:22 uh almost a similar increase, you know, 1/3 uh 1/3 decrease in like rollbacks, right? And so, I think people are writing code more productively and better code with it

57. EDGAR submissions index for Anthropic, PBC (CIK 2130357) · US Securities and Exchange Commission (EDGAR) · archived copy

    1. 59 EDGAR's submissions index for Anthropic, PBC (CIK 2130357), opened on 5 November 2026, lists a Form S-1 filed on 26 October 2026 and a Form S-1/A filed on 2 November 2026.
       checked 5 Nov 2026
       name Anthropic, PBC; filings.recent: form S-1/A, filingDate 2026-11-02, accessionNumber 0001628280-26-068676, primaryDocument anthropicpbc-sx1ax1.htm; form S-1, filingDate 2026-10-26, accessionNumber 0001628280-26-067168, primaryDocument anthropicpbc-sx1.htm

58. https://www.sec.gov/Archives/edgar/data/1181412/000162828026066482/spaceexplorationtechnologi.htm · archived copy

    1. 63 SpaceX's earnings release, Exhibit 99.1 to a Form 8-K (Item 2.02) filed with the SEC on 20 October 2026, reports results for the quarter ended 30 September 2026.
       Toned down Carriedfrom SpaceX reports $14.6 billion third-quarter revenue and a profit · checked 21 Oct 2026, not opened again for this piece
       Filing index 0001628280-26-066482, last modified 2026-10-20; Form 8-K Item 2.02: On October 20, 2026, Space Exploration Technologies Corp. issued an earnings release announcing results for the quarter ended September 30, 2026. Exhibit 99.1: SpaceX Reports Third Quarter 2026 Results
    2. 63 SpaceX reported revenues of $14.6 billion, up 161% from $5.6 billion.
       Carriedfrom SpaceX reports $14.6 billion third-quarter revenue and a profit · checked 21 Oct 2026, not opened again for this piece
       Revenues of $14.6 billion , up 161% from $5.6 billion
    3. 63 SI revenues were $7.6 billion, up 830% year-over-year, which SpaceX attributes to new cloud services agreements, the acquisition of Cursor, and increased Grok adoption.
       Carriedfrom SpaceX reports $14.6 billion third-quarter revenue and a profit · checked 21 Oct 2026, not opened again for this piece
       SI revenues were $7.6 billion , up 195% sequentially an d 830% year-over-year, primarily driven by the activation of new c loud services agreements , the acquisition of Cursor, and increased Grok adoption across enterprise customers and consumers

59. Space Exploration Technologies Corp. Form 8-K, Item 8.01 (filed 26 Oct 2026) · U.S. Securities and Exchange Commission · archived copy

    1. 64 SpaceX's Form 8-K, signed 26 October 2026, reports that on 24 October 2026 SpaceX entered into a binding commitment letter to enter into a master finance lease with an entity managed by Apollo Global Management, Inc.
       Carriedfrom SpaceX commits to about $40bn GPU lease with Apollo-managed lessor · checked 27 Oct 2026, not opened again for this piece
       On October 24, 2026, Space Exploration Technologies Corp. (the "Company") entered into a binding commitment letter ... an entity managed by Apollo Global Management, Inc. (such entity, the "Lessor"); signed October 26, 2026
    2. 64 The 8-K says SpaceX expects aggregate principal payments under the lease schedules over their terms to total approximately $40 billion.
       Carriedfrom SpaceX commits to about $40bn GPU lease with Apollo-managed lessor · checked 27 Oct 2026, not opened again for this piece
       aggregate principal payments under the lease schedules over their respective terms to total approximately $40 billion.
    3. 64 The 8-K says the lessor will acquire and lease to SpaceX graphics processing unit pods and related equipment for use at data centers selected by SpaceX.
       Carriedfrom SpaceX commits to about $40bn GPU lease with Apollo-managed lessor · checked 27 Oct 2026, not opened again for this piece
       the Lessor will acquire and lease to the Company certain graphics processing unit pods and related equipment (the "Equipment")

60. https://www.sec.gov/Archives/edgar/data/0001326801/000119312526432549/d382225dfwp.htm · archived copy

    1. 65 Meta's pricing term sheet gives a pricing date of October 26, 2026 for the Class A common stock offering and the Series A depositary shares offering.
       Carriedfrom Meta prices $25bn stock sale as Nvidia and AMD cite Muse AI agents · checked 30 Oct 2026, not opened again for this piece
       Pricing Term Sheet ... dated as of October 26, 2026 ... Pricing Date: October 26, 2026.
    2. 65 The term sheet says the common stock offering was upsized to 25.0 billion US dollars and the public offering price is 625.00 dollars per share, with settlement on October 29, 2026.
       Carriedfrom Meta prices $25bn stock sale as Nvidia and AMD cite Muse AI agents · checked 30 Oct 2026, not opened again for this piece
       Meta Platforms, Inc. has increased the size of the Common Stock Offering to $25.0 billion ... Public Offering Price of the Class A Common Stock: $625.00 per share of Class A Common Stock ... Settlement Date for the Common Stock Offering: October 29, 2026 (T+2)

61. https://nvidianews.nvidia.com/news/nvidia-and-meta-to-expand-collaboration-on-muse-and-muse-family-of-models-amid-robust-growth · archived copy

    1. 65 NVIDIA's release, dated October 26, 2026, says: In addition, NVIDIA will invest 3 billion dollars in Meta's equity offering. It gives no share count, price or valuation.
       Carriedfrom Meta prices $25bn stock sale as Nvidia and AMD cite Muse AI agents · checked 30 Oct 2026, not opened again for this piece
       October 26, 2026 ... In addition, NVIDIA will invest $3 billion in Meta’s equity offering.

62. NVIDIA and Supply Chain Partners Expand Capacity to Accelerate AI Infrastructure Buildout · NVIDIA · archived copy

    1. 66 The release states: "NVIDIA plans to increase substrate capacity for its products by 2.5x over the next three years."
       Carriedfrom Nvidia names six supply-chain partners to expand chip manufacturing capacity · checked 22 Oct 2026, not opened again for this piece
       NVIDIA plans to increase substrate capacity for its products by 2.5x over the next three years.

63. NVIDIA and Oracle Expand NVIDIA Vera Rubin to Multi-Gigawatt Scale and Bring NVIDIA Vera CPU to Oracle Cloud Infrastructure (joint release) · NVIDIA · archived copy

    1. 67 Oracle's and NVIDIA's joint release, dated October 27, 2026, says the companies "expanded their AI infrastructure collaboration with the general availability of NVIDIA Vera Rubin NVL72 and NVIDIA Vera CPUs on OCI."
       Carriedfrom Oracle says Nvidia Vera Rubin NVL72 racks are generally available, price not given · checked 30 Oct 2026, not opened again for this piece
       October 27, 2026 ... Oracle and NVIDIA also expanded their AI infrastructure collaboration with the general availability of NVIDIA Vera Rubin NVL72 and NVIDIA Vera CPUs on OCI.
    2. 67 The joint release states no price for the Vera Rubin NVL72 or Vera CPU capacity.
       Carriedfrom Oracle says Nvidia Vera Rubin NVL72 racks are generally available, price not given · checked 30 Oct 2026, not opened again for this piece
       (no price, no dollar figure anywhere in the release text)

64. Debtor motion for bidding procedures and Disney as stalking horse purchaser, Case 26-11603 (BLS), Doc 15 · US Bankruptcy Court, District of Delaware (Stretto case site) · archived copy

    1. 68 Character Technologies, Inc. filed a voluntary chapter 11 petition on 12 October 2026 (the Petition Date), Case No. 26-11603 (BLS), in the US Bankruptcy Court for the District of Delaware.
       Carriedfrom Disney agreed to hire most of Character's staff, filing says · checked 19 Oct 2026, not opened again for this piece
       Case No. 26-11603 (BLS) ... On October 12, 2026 (the "Petition Date"), the Debtor filed a voluntary petition for relief pursuant to chapter 11 of the Bankruptcy Code.
    2. 68 The filing says Disney will pay $20 million in cash under the Stalking Horse Transaction Agreement, of which $7.5 million was paid before the petition, including a $2 million refundable good-faith deposit.
       Carriedfrom Disney agreed to hire most of Character's staff, filing says · checked 19 Oct 2026, not opened again for this piece
       Disney agreed to pay, pursuant to the EPA, $7,500,000 of the $20,000,000 cash consideration up-front, $2,000,000 of which serves as a refundable good-faith deposit.
    3. 106 The proposed schedule in the filing sets a bid deadline of 20 November 2026 at 5:00 p.m. ET and an auction on 23 November 2026 at 10:00 a.m. ET.
       Carriedfrom Disney agreed to hire most of Character's staff, filing says · checked 19 Oct 2026, not opened again for this piece
       November 20, 2026 at 5:00 p.m. (ET) Bid Deadline ... the Auction shall be held on November 23, 2026 at 10:00 a.m. (ET)

65. Declaration of Adam Dunayer in Support of the Debtor's Bidding Procedures Motion, Case 26-11603 (BLS), Doc 17 · US Bankruptcy Court, District of Delaware (via CourtListener RECAP) · archived copy

    1. 68 The declaration in support of the motion (Doc 17) says Disney agreed to hire the majority of Character's workforce, including its chief executive officer, up front and before the Petition Date.
       Carriedfrom Disney agreed to hire most of Character's staff, filing says · checked 19 Oct 2026, not opened again for this piece
       Disney agreed to hire the majority of Character's workforce, including its chief executive officer, up front and prior to the Petition Date.

66. Anthropic Public Benefit Fund · Anthropic · archived copy

    1. 69 Anthropic published a post on 18 October 2026 about the Anthropic Public Benefit Fund.
       Carriedfrom Anthropic says it will commit 5% of its shares to a new Public Benefit Fund. · checked 19 Oct 2026, not opened again for this piece
       Investing in the public benefit ... Oct 18, 2026 ... Today, we're committing a portion of Anthropic's shares to the Anthropic Public Benefit Fund
    2. 69 Anthropic says that, subject to stockholder approval, 5% of its shares will be committed to the Anthropic Public Benefit Fund.
       Carriedfrom Anthropic says it will commit 5% of its shares to a new Public Benefit Fund. · checked 19 Oct 2026, not opened again for this piece
       Subject to stockholder approval, 5% of Anthropic's shares will be committed today

67. https://www.federalregister.gov/public-inspection/2026-21447/order-exclusion-and-removal-of-deepseek-products-from-federal-civilian-executive-branch-information · archived copy

    1. 70 The Department of Homeland Security published a notice in docket DHS-2026-0628 of an order from the Secretary of Homeland Security excluding DeepSeek (Hangzhou DeepSeek Artificial Intelligence Co., Ltd), including affiliated entities, from federal civilian agency procurements.
       Carriedfrom DHS orders DeepSeek products removed from civilian federal IT systems · checked 20 Oct 2026, not opened again for this piece
       The Department of Homeland Security is publishing notice of the Secretary of Homeland Security's issuance of an order (1) excluding information technology products and services developed by Hangzhou DeepSeek Artificial Intelligence Co., Ltd (DeepSeek), to include affiliated entities as defined in applicable regulations, from federal civilian agency procurements
    2. 70 The notice in docket DHS-2026-0628 is scheduled to be published in the Federal Register on 2026-10-20.
       Carriedfrom DHS orders DeepSeek products removed from civilian federal IT systems · checked 20 Oct 2026, not opened again for this piece
       This document is scheduled to be published in the Federal Register on 2026-10-20
    3. 71 The order says executive agency requests for exceptions must be submitted in writing within 60 days of the issuance of the order.
       Carriedfrom DHS orders DeepSeek products removed from civilian federal IT systems · checked 20 Oct 2026, not opened again for this piece
       Executive agency requests for exceptions must be submitted in writing within 60 days of the issuance of this order
    4. 90 The order applies to Federal Civilian Executive Branch Agencies except for their national security systems and secure compartmented information systems.
       Carriedfrom DHS orders DeepSeek products removed from civilian federal IT systems · checked 20 Oct 2026, not opened again for this piece
       Applies To: Federal Civilian Executive Branch Agencies except for their national security systems and secure compartmented information systems
    5. 100 The notice says the Secretary of Homeland Security ordered removal of all DeepSeek and DeepSeek-affiliate products, and third-party products and services developed using DeepSeek implementations, from civilian agency information technology.
       Carriedfrom DHS orders DeepSeek products removed from civilian federal IT systems · checked 20 Oct 2026, not opened again for this piece
       (2) removal of all DeepSeek and DeepSeek-affiliate products, and third-party products and services developed using DeepSeek implementations, from civilian agency information technology.

68. White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities (post with two attached pages) · Donald J. Trump on Truth Social · archived copy

    1. 72 Donald Trump posted "White House Accord on Super Intelligence - Joint Commitment on Frontier Responsibilities" on Truth Social on 29 September 2026, as a post with two attached images and no text of its own.
       Carriedfrom Six AI leaders sign voluntary commitment on frontier model controls · checked 5 Oct 2026, not opened again for this piece
       White House Accord on Super Intelligence / Joint Commitment on Frontier Responsibilities
    2. 72 The signature page lists Donald J. Trump, President of the United States, and Sundar Pichai, Google; Dario Amodei, Anthropic; Mark Zuckerberg, Meta; Greg Brockman, OpenAI; Elon Musk, XAI; and Jensen Huang, Nvidia.
       Carriedfrom Six AI leaders sign voluntary commitment on frontier model controls · checked 5 Oct 2026, not opened again for this piece
       Sundar Pichai Google / Dario Amodei Anthropic / Mark Zuckerberg Meta / Greg Brockman OpenAI / Elon Musk XAI / Jensen Huang Nvidia / Donald J. Trump President of the Unites States
    3. 73 The text says "we believe each company should implement the following four layers of controls and audits".
       Carriedfrom Six AI leaders sign voluntary commitment on frontier model controls · checked 5 Oct 2026, not opened again for this piece
       we believe each company should implement the following four layers of controls and audits
    4. 73 The text of the commitment states no penalty or enforcement mechanism.
       Carriedfrom Six AI leaders sign voluntary commitment on frontier model controls · checked 5 Oct 2026, not opened again for this piece
       Together, these steps will give each company, its customers, and the public confidence that the technology is operating as intended.

69. President Trump Participates in a Meeting and Luncheon on Super Intelligence · The White House · archived copy

    1. 74 Trump answered "Well, there's a belief that there should be tremendous self-regulation and we automatically have regulation with the Department of Justice, the FBI, all of that." (02:04).
       Carriedfrom Trump says "there's a belief" AI should have tremendous self-regulation · checked 4 Oct 2026, not opened again for this piece
       02:04 >> Well, there's a belief that there should / 02:06 be tremendous self-regulation and we / 02:08 automatically have regulation with the / 02:10 Department of Justice, the FBI, all of / 02:12 that.

70. Trump announces the formation of the Super Intelligence Force (SIF) · Donald J. Trump on Truth Social · archived copy

    1. 75 Donald Trump posted on Truth Social on 4 October 2026: "I am announcing the formation of the Super Intelligence Force (SIF)."
       Carriedfrom Trump announced a Super Intelligence Force on Truth Social on 4 October; the post gives no charter, budget or deadline. · checked 5 Oct 2026, not opened again for this piece
       I am announcing the formation of the Super Intelligence Force (SIF).
    2. 75 The text of the post names no charter, budget, deadline or legal authority for the Super Intelligence Force.
       Carriedfrom Trump announced a Super Intelligence Force on Truth Social on 4 October; the post gives no charter, budget or deadline. · checked 5 Oct 2026, not opened again for this piece
       The post's full text consists of the announcement, the task, the engagement list, the four leaders, the reporting line and a sign-off; no charter, budget, deadline or legal authority appears in it.

71. Petition for Panel Rehearing or Rehearing En Banc, Anthropic PBC v. U.S. Department of War, Nos. 26-1049 and 26-1162 (D.C. Cir., filed 13 October 2026), with the 25 September 2026 panel opinion in its addendum · United States Court of Appeals for the District of Columbia Circuit (filing by Anthropic's counsel) · archived copy

    1. 76 Anthropic filed a petition for panel rehearing or rehearing en banc in the U.S. Court of Appeals for the D.C. Circuit on 13 October 2026, in Nos. 26-1049 (lead) and 26-1162 (consolidated), Anthropic PBC v. U.S. Department of War.
       Carriedfrom Anthropic asks full D.C. Circuit to rehear Pentagon supply-chain ruling · checked 14 Oct 2026, not opened again for this piece
       No. 26-1049 (lead); No. 26-1162 (consolidated) UNITED STATES COURT OF APPEALS FOR THE DISTRICT OF COLUMBIA CIRCUIT ANTHROPIC PBC, Petitioner, v. U.S. DEPARTMENT OF WAR ... PETITION FOR PANEL REHEARING OR REHEARING EN BANC ... October 13, 2026
    2. 76 The petition asks the court to rehear the case after a divided panel decided it on 25 September 2026; its conclusion says only that the court should grant the petition.
       Carriedfrom Anthropic asks full D.C. Circuit to rehear Pentagon supply-chain ruling · checked 14 Oct 2026, not opened again for this piece
       A divided panel upheld that extraordinary designation. ... CONCLUSION The Court should grant the petition.
    3. 76 The opinion says the Department of War excluded Claude from its supply chain under the Federal Acquisition Supply Chain Security Act of 2018 after Anthropic refused to relax contractual prohibitions on the use of Claude for lethal autonomous warfare or domestic surveillance.
       Carriedfrom Anthropic asks full D.C. Circuit to rehear Pentagon supply-chain ruling · checked 14 Oct 2026, not opened again for this piece
       This case arises from a decision by the Department of War to exclude Claude, an artificialintelligence product developed by petitioner Anthropic PBC, from its supply chain under the Federal Acquisition Supply Chain Security Act of 2018. The Department made this decision after Anthropic refused to relax contractual prohibitions on the use of Claude for lethal autonomous warfare or domestic surveillance.

72. Anthropic PBC v. United States Department of War, No. 26-1049 (D.C. Cir.), docket · CourtListener (Free Law Project), from PACER

    1. 77 The CourtListener docket for Anthropic PBC v. United States Department of War, No. 26-1049 (D.C. Cir.), opened on 5 November 2026, lists a clerk's order of 3 November 2026 that vacates the order of 23 October 2026 and directs, on the court's own motion, that respondents file a response to Anthropic's petition for panel rehearing and rehearing en banc within 15 days.
       checked 5 Nov 2026
       Nov 3, 2026 CLERK'S ORDER [2201653] filed vacating the court's order filed October 23, 2026 [2199814-3]. It is FURTHER ORDERED, on the court's own motion, that within 15 days of the date of this order, respondents file a response to petitioner's petition for panel rehearing [2197902-3] and rehearing en banc [2197902-2].

73. Expanding our partnership with GSA to bring Claude to all levels of government · Anthropic · archived copy

    1. 78 Anthropic's post is dated Oct 27, 2026 and is titled "Expanding our partnership with GSA to bring Claude to all levels of government".
       Carriedfrom Anthropic offers governments 40% off Claude through a new GSA agreement · checked 28 Oct 2026, not opened again for this piece
       Expanding our partnership with GSA to bring Claude to all levels of government ... Published Oct 27, 2026
    2. 78 Anthropic says the new agreement "gives government partners a 40% discount on Claude Enterprise, Claude for Government, and the Claude API."
       Carriedfrom Anthropic offers governments 40% off Claude through a new GSA agreement · checked 28 Oct 2026, not opened again for this piece
       a new agreement that gives government partners a 40% discount on Claude Enterprise, Claude for Government, and the Claude API.
    3. 78 Anthropic says "Agencies are not bound by a minimum spend commitment and will be able to pay for the usage they consume."
       Carriedfrom Anthropic offers governments 40% off Claude through a new GSA agreement · checked 28 Oct 2026, not opened again for this piece
       Agencies are not bound by a minimum spend commitment and will be able to pay for the usage they consume.

74. Claude for Government is now generally available · Anthropic · archived copy

    1. 79 Anthropic says there are no seat fees: agencies pay for usage in fixed increments with a hard not-to-exceed cap.
       Carriedfrom Anthropic opens Claude for Government to federal and state agencies · checked 2 Oct 2026, not opened again for this piece
       No seat fees. Agencies pay for usage in fixed increments with a hard not-to-exceed cap, so spend does not exceed what an agency has obligated.

75. Government backs recommendations of NHS doctors-led AI Commission · GOV.UK · archived copy

    1. 80 The UK government said on 6 October 2026 that it will accept all 44 recommendations made by the National Commission into the Regulation of AI in Healthcare, published on 10 September. Quote: "The government has today, 6 October 2026, confirmed it will accept all 44 recommendations made by the National Commission into the Regulation of AI in Healthcare, published 10 September"
       Carriedfrom UK government accepts all 44 healthcare AI regulation recommendations · checked 7 Oct 2026, not opened again for this piece
       The government has today, 6 October 2026, confirmed it will accept all 44 recommendations made by the National Commission into the Regulation of AI in Healthcare, published 10 September
    2. 109 The first wave of AI Airlock Phase 3 innovators will be selected in November 2026, and a webinar for prospective applicants will be held on 22 October. Quote: "the first wave of innovators will be selected in November 2026. A webinar will be held for prospective applicants on 22 October."
       Carriedfrom UK government accepts all 44 healthcare AI regulation recommendations · checked 7 Oct 2026, not opened again for this piece
       AI Airlock Phase 3 applications are now open and the first wave of innovators will be selected in November 2026. A webinar will be held for prospective applicants on 22 October.

76. Building a Safer, More Secure Britain in the age of AI · GOV.UK · archived copy

    1. 81 Kanishka Narayan, the UK's Minister for Artificial Intelligence, gave the speech "Building a Safer, More Secure Britain in the age of AI" at Base One, Imperial College London, published on 12 October 2026.
       Carriedfrom UK minister says AI Security Institute is evaluating latest models · checked 13 Oct 2026, not opened again for this piece
       Building a Safer, More Secure Britain in the age of AI. Speech delivered by the Minister for Artificial Intelligence. From: Cabinet Office, Department for Business, Innovation, Science and Trade and The Rt Hon Kanishka Narayan MP. Published: 12 October 2026. Location: Base One, Imperial College London. Delivered on: 12 October 2026 (Transcript of the speech, exactly as it was delivered)
    2. 81 In the speech Narayan named Gemini 4 Argon among those models, saying the evaluations are "including Gemini 4 Argon to identify risks early".
       Carriedfrom UK minister says AI Security Institute is evaluating latest models · checked 13 Oct 2026, not opened again for this piece
       conducting rigorous evaluations of the latest generation of models, including Gemini 4 Argon to identify risks early, to test robust safeguards, to help keep people safe.

77. Code of Practice on marking and labelling of AI-generated content · European Commission · archived copy

    1. 82 The European Commission says the Article 50 transparency obligations are applicable from 2 August 2026, and that adherence to its Code of Practice on transparency of AI-generated content is voluntary while the Article 50 requirements are legal obligations.
       Carriedfrom How text watermarks work and what EU Article 50 requires · checked 15 Oct 2026, not opened again for this piece
       These transparency obligations, applicable from 2 August 2026 ... Even though adherence to the code is voluntary, the transparency requirements under article 50 of the AI Act are legal obligations.

78. Our approach to EU text provenance rules · OpenAI · archived copy

    1. 83 OpenAI says: "Over the coming weeks, we will add an invisible watermark to eligible ChatGPT and Codex text output in the European Union."
       Carriedfrom OpenAI will watermark ChatGPT and Codex text in the EU · checked 6 Oct 2026, not opened again for this piece
       Over the coming weeks, we will add an invisible watermark to eligible ChatGPT and Codex text output in the European Union.

79. Memorandum Opinion, Chegg v. Google and Penske Media v. Google (D.D.C.) · US District Court for the District of Columbia · archived copy

    1. 84 Judge Amit P. Mehta of the US District Court for the District of Columbia dated a memorandum opinion 30 September 2026 covering Chegg v. Google and Penske Media Corporation v. Google.
       Carriedfrom Judge dismisses Penske Media’s antitrust suit over Google AI Overviews · checked 2 Oct 2026, not opened again for this piece
       CHEGG, INC., Plaintiffs, v. GOOGLE LLC, et al. ... PENSKE MEDIA CORPORATION, et al., Plaintiffs, v. GOOGLE LLC, et al. ... MEMORANDUM OPINION ... Dated: September 30, 2026 Amit P. Mehta United States District Judge
    2. 84 The court granted Google’s motion to dismiss the Penske Media plaintiffs’ amended complaint.
       Carriedfrom Judge dismisses Penske Media’s antitrust suit over Google AI Overviews · checked 2 Oct 2026, not opened again for this piece
       Defendants' Motion to Dismiss the PMC Plaintiffs' Amended Complaint is also granted. PMC Docket, ECF No. 25.
    3. 84 The Penske Media plaintiffs also brought an unlawful tying claim, alleging that Google unlawfully ties AI Overviews to its general search product.
       Carriedfrom Judge dismisses Penske Media’s antitrust suit over Google AI Overviews · checked 2 Oct 2026, not opened again for this piece
       Finally, only the PMC Plaintiffs assert an unlawful tying claim in Count V of their Amended Complaint. ... They maintain that Google unlawfully ties AI Overviews to its general search product in violation of Section 2.

80. Complaint, ABG EPE IP LLC et al. v. Anthropic PBC, No. 5:26-cv-12473 (N.D. Cal., filed 23 Oct 2026), ECF 1 · U.S. District Court, N.D. Cal., via CourtListener RECAP · archived copy

    1. 85 The complaint, filed as Document 1 in Case No. 5:26-cv-12473 in the Northern District of California, is captioned as brought by ABG EPE IP LLC; The Luigi Creatore Trust; The Susan T. Aberbach Revocable Trust; The Decedents Exempt Trust created under the Anne-Rachel Aberbach Family Trust; Jason Simeon Aberbach; Abilene Music LLC; and Raleigh Music Publishing LLC against Anthropic PBC.
       Carriedfrom Anthropic sued over Claude output of 274 songs, some recorded by Elvis Presley · checked 27 Oct 2026, not opened again for this piece
       ABG EPE IP LLC; THE LUIGI CREATORE TRUST AND CLAIRE CREATORE AND STEPHEN CHAPMAN AS TRUSTEES; THE SUSAN T. ABERBACH REVOCABLE TRUST ... Case Number: 5:26-cv-12473
    2. 85 The complaint says the plaintiffs own or control exclusive rights in 274 musical compositions identified in its Exhibit A, including works recorded by Elvis Presley.
       Carriedfrom Anthropic sued over Claude output of 274 songs, some recorded by Elvis Presley · checked 27 Oct 2026, not opened again for this piece
       including some of the most timeless and iconic musical compositions in the world recorded by Elvis Presley. The 274 musical compositions are identified in the attached Exhibit A

81. Title 17, chapter 1: Section 107, limitations on exclusive rights: fair use · U.S. Copyright Office · archived copy

    1. 86 Section 107 lists four factors: the purpose and character of the use, including whether it is of a commercial nature or for nonprofit educational purposes; the nature of the copyrighted work; the amount and substantiality of the portion used in relation to the work as a whole; and the effect of the use upon the potential market for or value of the work.
       Carriedfrom What fair use is, and the four factors a US court weighs · checked 25 Oct 2026, not opened again for this piece
       the factors to be considered shall include— (1) the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes; (2) the nature of the copyrighted work; (3) the amount and substantiality of the portion used in relation to the copyrighted work as a whole; and (4) the effect of the use upon the potential market for or value of the copyrighted work.

82. We Must Pace the Frontier · darioamodei.com · archived copy

    1. 87 In “We Must Pace the Frontier” (September 2026) he proposes that each frontier AI company give embedded third-party evaluators “ongoing, employee-like access”, and writes that Anthropic is committing to this step now on its own; for Anthropic’s own external review team, he writes that reviewers should have the right to publish key findings “without editorial control by Anthropic”.
       Toned down Carriedfrom Who is Dario Amodei? · checked 13 Oct 2026, not opened again for this piece
       Each frontier AI company commits to giving ongoing, employee-like access to a team of embedded third-party evaluators (such as METR), whose role is to verify adherence to safety practices and commitments, report incidents, and help assess the alignment of not just completed AI models but training pipelines and processes. [...] Anthropic is unilaterally committing to this step now. [...] External reviewers should have the right to publish key findings about risk levels, incidents, practices, and the access they received or didn’t receive — without editorial control by Anthropic.

83. An update on independent evaluators · Anthropic · archived copy

    1. 88 Anthropic says it has onboarded six METR employees with company hardware and office space.
       Carriedfrom Anthropic onboards METR staff and cannot veto what they publish · checked 16 Oct 2026, not opened again for this piece
       We've onboarded six METR employees with company hardware, office space, and access to internal tools and systems
    2. 88 Anthropic says METR will publish its findings and that Anthropic cannot veto publication, but can require redaction of narrow categories of information.
       Carriedfrom Anthropic onboards METR staff and cannot veto what they publish · checked 16 Oct 2026, not opened again for this piece
       METR will publish its analysis of our risk levels, incidents, or practices. ... Anthropic can't veto publication, but we can require them to redact information ... if it falls into narrow, predefined categories

84. Altman on OpenAI’s IPO Plans, AI Agent Dots, AI Security · Bloomberg Television · archived copy

    1. 89 Altman answered that "when we need to make a decision about slowing down our training or our model release so that we can make more progress and put more of our attention into safety, alignment, monitoring, security, we will happily do that" (04:48).
       Carriedfrom Altman says OpenAI will slow training or release for safety · checked 3 Oct 2026, not opened again for this piece
       when we need to make a decision about slowing down our training or our model release so that we can make more progress and put more of our attention into safety, alignment, monitoring, security, we will happily do that.

85. How employment is changing at Anthropic · Anthropic · archived copy

    1. 91 The report "How employment is changing at Anthropic" was published on 16 October 2026 as a technical report from The Anthropic Institute.
       Carriedfrom Anthropic says junior technical hires fell as agent work grew · checked 19 Oct 2026, not opened again for this piece
       How employment is changing at Anthropic ... Oct 16, 2026 ... @techreport{wilmers2026employment, ... institution = {The Anthropic Institute}, type = {Technical Report}
    2. 91 The report says the junior share of technical hires fell from 27% to 9%.
       Carriedfrom Anthropic says junior technical hires fell as agent work grew · checked 19 Oct 2026, not opened again for this piece
       The proportion of junior technical hires fell even faster, from 27% to 9%.
    3. 91 The report says agent work passed human work hours in March 2026 and that by August interactive agent hours were about six times employee hours.
       Carriedfrom Anthropic says junior technical hires fell as agent work grew · checked 19 Oct 2026, not opened again for this piece
       in March 2026, the sum of autonomous and interactive agent work hours leapt past human work hours for the first time. ... By August interactive AI agent hours outnumbered employees' six to one.
    4. 92 The report compares 192 engineers who adopted coding agents with 74 who did not, and says tool adoption was not random, so the estimates may be biased.
       Carriedfrom Anthropic says junior technical hires fell as agent work grew · checked 19 Oct 2026, not opened again for this piece
       Note, however, that tool adoption is not random: if employees on the cusp of driving some big new project decide to tool up for it, for example, this would bias our estimates. We follow 192 engineers over time as they adopt coding agents, compared to 74 engineers who do not adopt agents during the period studied.

86. It Does the Work and Doesn't Cry: Ethan Mollick on AI Replacing Interns (The Prof G Pod) · The Prof G Pod - Scott Galloway (YouTube) · archived copy

    1. 93 Mollick said "middle managers were increasingly turning to using AI instead of interns, cuz it it does the work and doesn't cry" (02:24).
       Carriedfrom Wharton's Mollick says managers now give interns' work to AI · checked 12 Oct 2026, not opened again for this piece
       02:24 than you at your job. And middle / 02:26 managers were increasingly turning to / 02:27 using AI instead of interns, cuz it it / 02:29 does the work and doesn't cry, right?

87. AI Agents Cut 3-Day Portfolio Builds to Minutes - Jack McDonald, Apollo & Michael Manapat, Rowspace · AI Engineer (YouTube) · archived copy

    1. 94 Jack McDonald introduced himself: 'I am a principal at Apollo and COO of Reddington Ridge Capital' (00:55); the automatic captions render Redding Ridge as 'Reddington Ridge'.
       Carriedfrom Apollo's McDonald says AI cut mock portfolio builds to minutes · checked 18 Oct 2026, not opened again for this piece
       00:53 Yeah, thank you, Michael. Uh I'm Jack 00:55 McDonald. I am a principal at Apollo and 00:58 COO of Reddington Ridge Capital
    2. 94 McDonald said building such a portfolio would typically take a portfolio manager before adopting Rowspace 'anywhere from two to three days in a manual Excel' (04:44); the captions render Rowspace as 'Rose special'.
       Carriedfrom Apollo's McDonald says AI cut mock portfolio builds to minutes · checked 18 Oct 2026, not opened again for this piece
       04:40 would typically take a PM before kind of 04:44 adopting Rose special anywhere from two 04:46 to three days in a manual Excel.
    3. 94 McDonald said the team can now create a portfolio 'that would take somebody days to now do it in the matter of minutes' (04:53).
       Carriedfrom Apollo's McDonald says AI cut mock portfolio builds to minutes · checked 18 Oct 2026, not opened again for this piece
       04:53 create a portfolio that would take 04:55 somebody days to now do it in the matter 04:57 of minutes.

88. Cutting our inference bill by 68% by being obsessive about optimization · Cloudflare Blog · archived copy

    1. 95 Cloudflare's engineering blog post of 19 October 2026 says that with code quality at high effort the average review of its AI code reviewer now costs $0.77.
       Carriedfrom Cloudflare says its AI code reviews now cost 68% less · checked 20 Oct 2026, not opened again for this piece
       October 19, 2026 Cutting our inference bill by 68% by being obsessive about optimization ... With code quality at high effort, the average review now costs just $0.77
    2. 95 The post says that $0.77 is down 68% from the $2.43 the average review cost over its two most expensive weeks before the work started.
       Carriedfrom Cloudflare says its AI code reviews now cost 68% less · checked 20 Oct 2026, not opened again for this piece
       the average review now costs just $0.77, down 68% from the $2.43 it averaged over our two most expensive weeks before we started.
    3. 95 The post says the median review now takes 1m 32s instead of the 4m 09s it took before.
       Carriedfrom Cloudflare says its AI code reviews now cost 68% less · checked 20 Oct 2026, not opened again for this piece
       the median review now takes just 1m 32s instead of the 4m 09s it took before.

89. This AI Expert Argues We Are Underestimating AI Impact — With Ethan Mollick · Big Technology Podcast (YouTube) · archived copy

    1. 96 Ethan Mollick said a company he spoke to went from a $200 million data migration plan for its old mainframe data to a $20 million plan a few months ago, and its CTO is thinking they may do it for $2 million in a couple of months (14:20). The company is not named.
       Carriedfrom Mainframe migration quote fell from $200 million to $20 million, Mollick recounts · checked 31 Oct 2026, not opened again for this piece
       (14:18) Like I just spoke to a company, they went from a $200 million data migration plan for their old, you know, old mainframe data to they renegotiate to a $20 million plan just a few months ago, and the CTO is thinking like, well, maybe I can do this for $2 million in a couple months.

90. Gemini 4 Argon: Built to solve your hardest problems · Google · archived copy

    1. 97 Google's post of 13 October 2026 names the model Gemini 4 Argon, and says that inside Google, Argon in Antigravity has helped teams increase agentic code submissions by more than 35% while reducing code rollbacks by nearly a third; these are Google's own figures about its own engineers.
       Toned down Carriedfrom Pichai says Google's rollbacks fell a third after Gemini 4 Argon launched · checked 25 Oct 2026, not opened again for this piece
       At Google, we’ve been using Argon on our hardest engineering problems. Over the past few weeks, software engineers have been shipping high-quality code faster than ever. Argon in Antigravity has helped teams increase agentic code submissions by more than 35% while reducing code rollbacks by nearly a third.

91. Leadership - SAP Executive Board · SAP · archived copy

    1. 98 SAP's leadership page lists Christian Klein as Chief Executive Officer (CEO) of SAP.
       Carriedfrom SAP's Klein says agents may reach 100% on some tasks · checked 5 Oct 2026, not opened again for this piece
       Christian Klein Chief Executive Officer (CEO)

92. SAP CEO: AI Won't Kill Software, But It Will Change Your Job — With Christian Klein · Alex Kantrowitz (Big Technology Podcast) · archived copy

    1. 98 Klein said: "now the token spend is really coming to a point where we definitely need to set certain limits" (48:46).
       Carriedfrom SAP's Klein says agents may reach 100% on some tasks · checked 5 Oct 2026, not opened again for this piece
       48:46 now the token spend is really coming to / 48:48 a point where we definitely need to / 48:51 set certain limits.
    2. 98 Klein spoke of "tone spend", the automatic captions' rendering of token spend: "we have a approval process where a manager can also say okay for this month or for this task or for this project we give you more tone spend" (49:53).
       Carriedfrom SAP's Klein says agents may reach 100% on some tasks · checked 5 Oct 2026, not opened again for this piece
       49:53 And then we have a approval process / 49:55 where a manager can also say okay for / 49:57 this month or for this task or for this / 49:59 project we give you more tone spend.

93. AL26-026 - Critical vulnerabilities impacting PHP - CVE-2026-109618 and CVE-2026-109619 · Canadian Centre for Cyber Security · archived copy

    1. 99 Alert AL26-026 says: "All PHP installations serving web requests, including PHP-FPM, Apache with mod_php, php-cgi, FrankenPHP and LiteSpeed are at risk."
       Carriedfrom Critical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers · checked 23 Oct 2026, not opened again for this piece
       All PHP installations serving web requests, including PHP-FPM, Apache with mod_php, php-cgi, FrankenPHP and LiteSpeed are at risk.

94. 2026 Usage Policy update · Anthropic · archived copy

    1. 105 Anthropic says the updated Usage Policy takes effect on 12 November.
       Carriedfrom Anthropic's updated Usage Policy will take effect on 12 November · checked 9 Oct 2026, not opened again for this piece
       The updated policy takes effect on November 12.
    2. 105 The updated policy, which takes effect on 12 November 2026, adds requirements for Claude connected to hardware that takes autonomous physical actions and might cause injury: a qualified operator must be able to observe the equipment and stop it, and the equipment must be able to hold a safe state if Claude is disconnected.
       Toned down Carriedfrom Anthropic's updated Usage Policy will take effect on 12 November · checked 9 Oct 2026, not opened again for this piece
       this section also adds requirements for the use of our models when they're connected to hardware that takes autonomous physical actions and which might be capable of causing injury. A qualified operator must be able to observe the equipment and stop it if needed; the equipment must also be able to hold a safe state if Claude is disconnected.

95. Character Technologies, Inc., No. 26-11603 (Bankr. D. Del.), docket · CourtListener (Free Law Project), from PACER

    1. 107 The CourtListener docket for Character Technologies, Inc., No. 26-11603 (Bankr. D. Del.), opened on 5 November 2026 and showing entries through 4 November 2026, lists no entry described as an order approving the bidding procedures; six entries of 4 November 2026 (106 and 108 to 112) are listed only as Order.
       checked 5 Nov 2026
       106 Nov 4, 2026 Order ... 108 Nov 4, 2026 Order ... 109 Nov 4, 2026 Order ... 110 Nov 4, 2026 Order ... 111 Nov 4, 2026 Order ... 112 Nov 4, 2026 Order ... 115 Nov 4, 2026 Notice of Appointment of Creditors' Committee

96. Gemini Developer API pricing (ai.google.dev/gemini-api/docs/pricing redirects here) · Google · archived copy

    1. 108 Google's Gemini API pricing page, opened on 5 November 2026, lists Gemini 4 Argon's Standard paid-tier prices per million tokens as $2.00 input and $10.00 output through November 30, 2026, and $4.00 input and $20.00 output starting December 1, 2026; the page does not mention Gemini 4 Carbon.
       checked 5 Nov 2026
       Gemini 4 Argon ... Standard ... Paid Tier, per 1M tokens in USD ... Input price ... $2.00 through November 30, 2026. $4.00 starting December 1, 2026. ... Output price (including thinking tokens) ... $10.00 through November 30, 2026. $20.00 starting December 1, 2026.

97. Preparing for the 2026 US midterms · Anthropic · archived copy

    1. 110 Anthropic published a post titled "Preparing for the 2026 US midterms" dated 14 October 2026.
       Carriedfrom Anthropic sets out its safeguards for the 2026 US midterms · checked 15 Oct 2026, not opened again for this piece
       Preparing for the 2026 US midterms ... Oct 14, 2026
    2. 110 Anthropic says it has added new classifiers to identify novel and multi-step misuse, including attempts to build fake social media accounts and news sites aimed at political discussions, or to run financially motivated election schemes.
       Carriedfrom Anthropic sets out its safeguards for the 2026 US midterms · checked 15 Oct 2026, not opened again for this piece
       we've added new classifiers to better identify novel and multi-step misuse, including attempts to build fake social media accounts and news sites aimed at political discussions, or to run financially motivated election schemes.