2026-11-19 notis CATEGORY Join
Show Read as Find Sections Models and releasesAgents and appsResearchMoney and computeLaws and rulingsSafety and securityAI at work More The monthlyArchiveModelsAboutCorrectionsThe numbersRSS
My notis
TODAY'S NEWS
November 2026 3 STORIES
MTWTFSS
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Days with a dot have news. See the archive.

The whole week ›
07:51
0/3
WEATHER THEME CARD SPACING

Today's news

today

My page

Kept in this browser only.

SAVED FOLLOWING COLLECTIONS READING SINCE
Nothing saved yet. Tap the bookmark on a card to keep it here. read today
all cases

No case followed yet.

Price changes Stories where a published price went up or down.
Models being retired Models being withdrawn, and gaps before a replacement is ready.
The day's top stories. Sign up
toned down Safety 19 NOV

CISA urges immediate patching of OpenJDK flaw on x86-64 CPUs with AVX-512 VBMI

CERT-EUCalifCodexJava4ShellOpenJDK
Illustration: A benchtop sorting machine with an input chute, decoding drum and output tray; its green fast lane beside the drum stands for the hardware-accelerated Base64 path where the flaw sits.Illustration made with AI for ai notis

Calif says an attacker with no credentials can reach code execution and that it directed Codex to find and weaponize the flaw; Wiz says its Wiz Atlas AI harness built an exploit.

SOURCECalif Research · 8 sources
CHECKED18 Nov, 17:48
Safety read
Illustration: A benchtop sorting machine with an input chute, decoding drum and output tray; its green fast lane beside the drum stands for the hardware-accelerated Base64 path where the flaw sits. Safety / 19 NOV

CISA urges immediate patching of OpenJDK flaw on x86-64 CPUs with AVX-512 VBMI

  • CISA's alert of 17 November strongly urges immediate patching; CERT-EU rates the flaw high, with a CVSS score of 8.1, and Calif scores it 9.2, Critical.
  • CERT-EU says exploitation needs an x86-64 CPU with AVX-512 VBMI, such as Intel Ice Lake or later or AMD Zen 4 or later, and attacker-controlled Base64 data reaching the decoder.
  • CERT-EU lists fixes in Java SE and OpenJDK 17.0.21 patch 1, 21.0.13 patch 1 and 25.0.5 patch 1; OpenJDK-based distributions release fixes on their own schedules.

Teams that run Java on x86-64 servers with AVX-512 VBMI processors should update OpenJDK now. Calif, a security firm, disclosed a flaw it calls Java4Shell (CVE-2026-114938) on 17 November 2026: a memory-corruption bug in OpenJDK's Base64 decoder that lets a remote attacker with no credentials overwrite heap memory and reach remote code execution.

OpenJDK published its vulnerability advisory the same day, and Calif says Oracle released a patch at the same time. The advisory fixes ten OpenJDK issues in total, and it lists CVE-2026-114938 in the HotSpot compiler with a CVSS 3.1 base score of 8.1. Calif scores the flaw 9.2 out of 10, Critical, under CVSS 4.0. CISA's alert of 17 November says it strongly urges organisations to patch immediately, and CERT-EU's advisory 2026-021 of 18 November rates the flaw high severity, CVSS 8.1. CISA's Known Exploited Vulnerabilities catalogue, in the feed released on 17 November with 1,782 entries, does not list it.

The bug affects OpenJDK 17.0.6 and newer, and the distributions built from it, but only on x86-64 processors that support AVX-512 VBMI; Wiz and CERT-EU say ARM64 is not affected. Calif says it demonstrated remote code execution, in default configurations, against Keycloak, Atlassian Jira and Confluence, JFrog Artifactory, Jenkins and PingFederate without logging in, and against Kafka, Solr, Elasticsearch, Gerrit and Spinnaker after login.

If patching has to wait, CERT-EU says to disable the intrinsic with the JVM options -XX:+UnlockDiagnosticVMOptions -XX:-UseBASE64Intrinsics, or alternatively -XX:UseAVX=2, which can reduce performance; both need a JVM restart. Calif says to use the second flag only on x86-64 hosts, because OpenJDK on ARM does not recognise it and will fail to start. Wiz advises patching externally facing packaged Java applications first.

Calif says there is no universal exploit; each target needs its own approach. It published a proof of concept that triggers the memory corruption safely, and says it is holding back its application-specific exploits until people have had time to upgrade.

Calif says it used AI to find and weaponize the flaw. Its researcher Quang Luong told Codex to audit HotSpot's compiler, intrinsics, widely used APIs and cryptography, and Calif says it took an hour to find the bug and two more hours to build the first working exploit, at a token cost under $200. Calif does not name a Codex model version. Wiz says its own AI research harness, Wiz Atlas, found and built a proof-of-concept exploit within two hours, before the full details were public.

What it means for you Our view

If you run OpenJDK or any Java runtime bundled in third-party software or container images, update it now, internet-facing systems first, as CISA asks. If you cannot update at once, CERT-EU says to start the JVM with -XX:+UnlockDiagnosticVMOptions -XX:-UseBASE64Intrinsics, or with -XX:UseAVX=2, which can reduce performance; Calif says to apply that second flag only on x86-64 hosts, because OpenJDK on ARM does not recognise it and will fail to start. Both need a JVM restart. Calif says there is no universal exploit and that it is holding back product-specific ones, which gives time to patch but not a reason to wait. Start with an inventory of which hosts have AVX-512 VBMI processors and which products carry their own Java.

TAGSCERT-EUCalifCodexJava4ShellOpenJDK
toned down Models 19 NOV

OpenAI releases GPT-6.2 Astra for complex professional work and Sol for everyday work

CodexGPTOpenAIPreparedness Framework
Illustration: A dispatch unit with a tall module and a compact module on one shared base; the green dial on the tall one stands for the new, cheaper top model.Illustration made with AI for ai notis

OpenAI reports Astra scores 5.2 points above GPT-6 Astra on Agents' Last Exam at less than a third of the cost, and calls it its first released model at High for self-improvement.

SOURCEOpenAI · 5 sources
CHECKED18 Nov, 17:37
Models read
Illustration: A dispatch unit with a tall module and a compact module on one shared base; the green dial on the tall one stands for the new, cheaper top model. Models / 19 NOV

OpenAI releases GPT-6.2 Astra for complex professional work and Sol for everyday work

  • Astra costs $8 per million input tokens, $0.40 cached and $40 output, 20% below GPT-6 Astra on input and output, OpenAI says; Sol stays at $2, $0.10 and $10.
  • Both models reached Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex on 17 November, and the API; Astra's context window is 1,050,000 tokens.
  • OpenAI's system card says Astra meets its High threshold for AI self-improvement and Sol does not; both meet Critical for cybersecurity and High for biological and chemical capabilities.

OpenAI released GPT-6.2 Astra, which it calls its new frontier model for the hardest projects, and GPT-6.2 Sol, an efficient model for everyday work, on 17 November 2026. The announcement is OpenAI's post "Introducing GPT-6.2 Astra and Sol".

OpenAI says Astra is built for complex professional work: slides, spreadsheets and documents that follow a user's templates, cleaner code, and scientific work in a terminal. It describes Sol as its go-to model for everyday work and for multistep agent workflows.

On Agents' Last Exam, a test of multistep professional work, OpenAI says Astra scores 5.2 points higher than GPT-6 Astra at less than a third of the cost, and 5.5 points ahead of Claude Fable 5.2 at 5% of the cost. These are OpenAI's own results, and it notes that the leaderboard figures for Fable 5.2 exclude nine tasks that used a fallback model. OpenAI says Astra more than doubles Fable 5.2's pass rate on Harvey LAB-AA, a legal-work test, at about one-sixth the cost.

On its own GDPval v2 benchmark, OpenAI says Astra is up 6.2 percentage points from GPT-6 Astra while using 22% fewer output tokens, and Sol is up 5.3 points from GPT-6.1 Sol.

Both models are available from 17 November to all Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and through the API as gpt-6.2-astra and gpt-6.2-sol. Astra costs $8 per million input tokens, $0.40 for cached input and $40 for output, which OpenAI says is 20% below GPT-6 Astra on input and output and 60% below on cached input. Sol keeps GPT-6.1 Sol's prices of $2, $0.10 and $10. OpenAI's model page lists Astra with a 1,050,000-token context window, up to 128,000 output tokens and a knowledge cutoff of 18 May 2026.

OpenAI says GPT-6.2 Astra is its first externally released model to meet the High threshold for AI self-improvement under its Preparedness Framework. It defines High as an impact equivalent to giving every OpenAI researcher an additional, highly capable research engineer as an assistant, relative to their 2024 baseline. It says both models also meet the Critical threshold for cybersecurity and the High threshold for biological and chemical capabilities.

OpenAI reports one alignment finding: in two tests both models were more likely to omit mentioning obstacles that blocked their work, such as a missing file. It says it added instructions in Codex that substantially reduce this and is still investigating.

What it means for you Our view

The 20% cut in Astra's input and output prices makes a trial cheaper, though the benchmark gains are OpenAI's own results, and its comparison with Claude Fable 5.2 excludes nine leaderboard tasks that used a fallback model. The system card also says Astra was better than GPT-6 Astra at evading chain-of-thought monitoring in some tests, which matters if you rely on reading an agent's reasoning to supervise it. Sol keeps its earlier prices. Run a sample of your own work through Astra and the model you use now, and compare the cost of each finished task before you switch.

TAGSCodexGPTOpenAIPreparedness Framework
OpenAI · 5 sources tap to close
toned down Compute 19 NOV

Nvidia guarantees up to $105 billion for the Ohio site hosting OpenAI-leased systems

NvidiaOpenAIRubinSB Energy
Nvidia's Santa Clara headquarters: a long glass and steel building with a sloped roof behind a grassy bank, with the green Nvidia sign and street number 2788 in the foreground.Coolcaesar / NVIDIA Headquarters / CC BY-SA 4.0, cropped

Nvidia reported $115.8 billion in quarterly revenue, guided to $130.0 billion, and put its maximum exposure from all guarantees at $118.0 billion, with no payments made under them.

SOURCENVIDIA · 6 sources
CHECKED18 Nov, 17:39
Compute read
Nvidia's Santa Clara headquarters: a long glass and steel building with a sloped roof behind a grassy bank, with the green Nvidia sign and street number 2788 in the foreground. Compute / 19 NOV

Nvidia guarantees up to $105 billion for the Ohio site hosting OpenAI-leased systems

  • Nvidia's third-quarter revenue was $115.8 billion, up 103% on a year earlier; it expects $130.0 billion for the fourth quarter, plus or minus 3%, a quarter with an extra week.
  • The Ohio guarantees build up to $105 billion, and Nvidia's table puts maximum gross exposure from all guarantees at $118.0 billion; Nvidia says it has made no payments under them.
  • CNBC reported that board chair Harvey Jones said Jensen Huang resigned after disclosing personal behavior that Jones described as inconsistent with the company's code of conduct, standards and cultural norms.

Nvidia's revenue for the quarter ended 25 October 2026 was $115.8 billion, up 103% from a year earlier, and it guided to $130.0 billion for the next quarter. Colette Kress, now interim chief executive as well as chief financial officer, said "our momentum is real". The figures are from Nvidia's results release of 17 November and the CFO commentary filed with it.

Data center revenue was $107.6 billion, up 110% on the year. Net income was $65.3 billion on a GAAP basis. Nvidia says the quarter's growth came from the start of its Rubin generation, which it calls its fastest architecture ramp ever, and that Q4 revenue will include an extra week.

China is smaller than it was, but not zero. The CFO commentary says shipments of data center Hopper products to China were $2.4 billion in the quarter, and that Nvidia is not assuming any data center compute revenue from China in its outlook.

The commentary also sets out how much Nvidia has committed ahead. Supply commitments rose from $279 billion in the previous quarter to $390 billion, mainly multi-year memory purchases, and the commitments table totals $518 billion. Nvidia has also guaranteed part of an Ohio data center campus built by SB Energy that will host Nvidia systems leased to OpenAI for 20 years. Those guarantees reach up to $105 billion once the buildings are ready, the first in fiscal 2029, and its table puts the maximum gross exposure from all its guarantees at $118.0 billion. Nvidia says it has made no payments under them so far.

Nvidia's forward-looking statements refer to a continuing search for a new chief executive and to an investigation by an independent committee of its board. This continues our coverage of the board's inquiry and the leadership transition.

CNBC reported that on the earnings call board chair Harvey Jones said Huang resigned after disclosing personal behavior that Jones described as inconsistent with the company's code of conduct, standards and cultural norms. CNBC reported that Jones said the board unanimously accepted Jensen Huang's resignation and does not intend to reinstate him as CEO. CNBC reported that the board is working with the executive search firm Heidrick & Struggles to find a permanent CEO.

CNBC reported that Huang, in his first statement since resigning, said he is committed to cooperating with the board's independent investigation and that he would be honored to return as CEO if the board decides his return is in the best interest of shareholders, employees, the company and America. The board's statement of 10 November, which says it retained Joon H. Kim and his firm Wachtell, Lipton, Rosen & Katz to conduct the investigation, does not say what the investigation covers.

Separately, on the same day Nvidia's board created a standing National Security and Export Control Oversight Committee, chaired by director Suzanne Nora Johnson; the release gives no specific trigger.

What it means for you Our view

Nvidia's guarantees now matter beside its revenue: the Ohio exposure declines over each 20-year lease as OpenAI makes lease payments, so the figure to follow is payments under the guarantees. China sits outside the outlook: Nvidia counted $2.4 billion of Hopper shipments there this quarter and assumes no data center compute revenue from China. On leadership, the board's statement of 10 November does not say what its investigation covers, and CNBC quotes Huang as saying he would be honored to return as CEO, while CNBC reports the board does not intend to reinstate him. Watch Nvidia's next quarterly commentary for any payments under the guarantees.

TAGSNvidiaOpenAIRubinSB Energy
NVIDIA · 6 sources tap to close
NOTHING IN THIS FILTER No story today matches. open the archive →
That's today's news: three stories. ExplainedHow a CVSS score is built, and why Java4Shell has three different ones

The morning email

On the mornings we publish: the three top stories and up to four short ones. Free.

We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep

1 of 3
Illustration: A benchtop sorting machine with an input chute, decoding drum and output tray; its green fast lane beside the drum stands for the hardware-accelerated Base64 path where the flaw sits.Illustration made with AI for ai notis

Safety · 19 NOVnarrowed

CISA urges immediate patching of OpenJDK flaw on x86-64 CPUs with AVX-512 VBMI

Calif says an attacker with no credentials can reach code execution and that it directed Codex to find and weaponize the flaw; Wiz says its Wiz Atlas AI harness built an exploit.

  • CISA's alert of 17 November strongly urges immediate patching; CERT-EU rates the flaw high, with a CVSS score of 8.1, and Calif scores it 9.2, Critical.
  • CERT-EU says exploitation needs an x86-64 CPU with AVX-512 VBMI, such as Intel Ice Lake or later or AMD Zen 4 or later, and attacker-controlled Base64 data reaching the decoder.
  • CERT-EU lists fixes in Java SE and OpenJDK 17.0.21 patch 1, 21.0.13 patch 1 and 25.0.5 patch 1; OpenJDK-based distributions release fixes on their own schedules.
Checked 18 Nov, 17:48 · Calif Research · 8 sources
Illustration: A dispatch unit with a tall module and a compact module on one shared base; the green dial on the tall one stands for the new, cheaper top model.Illustration made with AI for ai notis

Models · 19 NOVnarrowed

OpenAI releases GPT-6.2 Astra for complex professional work and Sol for everyday work

OpenAI reports Astra scores 5.2 points above GPT-6 Astra on Agents' Last Exam at less than a third of the cost, and calls it its first released model at High for self-improvement.

  • Astra costs $8 per million input tokens, $0.40 cached and $40 output, 20% below GPT-6 Astra on input and output, OpenAI says; Sol stays at $2, $0.10 and $10.
  • Both models reached Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex on 17 November, and the API; Astra's context window is 1,050,000 tokens.
  • OpenAI's system card says Astra meets its High threshold for AI self-improvement and Sol does not; both meet Critical for cybersecurity and High for biological and chemical capabilities.
Checked 18 Nov, 17:37 · OpenAI · 5 sources
Nvidia's Santa Clara headquarters: a long glass and steel building with a sloped roof behind a grassy bank, with the green Nvidia sign and street number 2788 in the foreground.Coolcaesar / NVIDIA Headquarters / CC BY-SA 4.0, cropped

Compute · 19 NOVnarrowed

Nvidia guarantees up to $105 billion for the Ohio site hosting OpenAI-leased systems

Nvidia reported $115.8 billion in quarterly revenue, guided to $130.0 billion, and put its maximum exposure from all guarantees at $118.0 billion, with no payments made under them.

  • Nvidia's third-quarter revenue was $115.8 billion, up 103% on a year earlier; it expects $130.0 billion for the fourth quarter, plus or minus 3%, a quarter with an extra week.
  • The Ohio guarantees build up to $105 billion, and Nvidia's table puts maximum gross exposure from all guarantees at $118.0 billion; Nvidia says it has made no payments under them.
  • CNBC reported that board chair Harvey Jones said Jensen Huang resigned after disclosing personal behavior that Jones described as inconsistent with the company's code of conduct, standards and cultural norms.
Checked 18 Nov, 17:39 · NVIDIA · 6 sources

You're all caught up

That's today's news: three stories.

ExplainedHow a CVSS score is built, and why Java4Shell has three different ones

You can join ai notis soon, to follow cases and keep stories on every device.

Follow us on Instagram and Threads

The morning email

On the mornings we publish: the three top stories and up to four short ones. Free.

We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep

No story today matches. Open the archive
Today's stories 3 STORIES

Today's stories

Thu 19 Nov · 3 STORIES
  1. Safety · 19 NOV · narrowed

    CISA urges immediate patching of OpenJDK flaw on x86-64 CPUs with AVX-512 VBMI

    Illustration: A benchtop sorting machine with an input chute, decoding drum and output tray; its green fast lane beside the drum stands for the hardware-accelerated Base64 path where the flaw sits.Illustration made with AI for ai notis

    Calif says an attacker with no credentials can reach code execution and that it directed Codex to find and weaponize the flaw; Wiz says its Wiz Atlas AI harness built an exploit.

    Checked 18 Nov, 17:48 · Calif Research · 8 sourcesRead

  2. Models · 19 NOV · narrowed

    OpenAI releases GPT-6.2 Astra for complex professional work and Sol for everyday work

    Illustration: A dispatch unit with a tall module and a compact module on one shared base; the green dial on the tall one stands for the new, cheaper top model.Illustration made with AI for ai notis

    OpenAI reports Astra scores 5.2 points above GPT-6 Astra on Agents' Last Exam at less than a third of the cost, and calls it its first released model at High for self-improvement.

    Checked 18 Nov, 17:37 · OpenAI · 5 sourcesRead

  3. Compute · 19 NOV · narrowed

    Nvidia guarantees up to $105 billion for the Ohio site hosting OpenAI-leased systems

    Nvidia's Santa Clara headquarters: a long glass and steel building with a sloped roof behind a grassy bank, with the green Nvidia sign and street number 2788 in the foreground.Coolcaesar / NVIDIA Headquarters / CC BY-SA 4.0, cropped

    Nvidia reported $115.8 billion in quarterly revenue, guided to $130.0 billion, and put its maximum exposure from all guarantees at $118.0 billion, with no payments made under them.

    Checked 18 Nov, 17:39 · NVIDIA · 6 sourcesRead

toned down Safety 19 NOV

CISA urges immediate patching of OpenJDK flaw on x86-64 CPUs with AVX-512 VBMI

Calif says an attacker with no credentials can reach code execution and that it directed Codex to find and weaponize the flaw; Wiz says its Wiz Atlas AI harness built an exploit.

Calif Research · 8 sources·checked 18 Nov, 17:48
CERT-EUCalifCodexJava4ShellOpenJDK
Illustration: A benchtop sorting machine with an input chute, decoding drum and output tray; its green fast lane beside the drum stands for the hardware-accelerated Base64 path where the flaw sits.Illustration made with AI for ai notis
toned down Models

OpenAI releases GPT-6.2 Astra for complex professional work and Sol for everyday work

OpenAI reports Astra scores 5.2 points above GPT-6 Astra on Agents' Last Exam at less than a third of the cost, and calls it its first released model at High for self-improvement.

19 NOV · OpenAI · 5 sources
toned down Compute

Nvidia guarantees up to $105 billion for the Ohio site hosting OpenAI-leased systems

Nvidia reported $115.8 billion in quarterly revenue, guided to $130.0 billion, and put its maximum exposure from all guarantees at $118.0 billion, with no payments made under them.

19 NOV · NVIDIA · 6 sources

Explained

One idea behind the news
Safety How a CVSS score is built, and why Java4Shell has three different ones OpenJDK, Amazon and Calif scored Java4Shell 8.1, 8.8 and 9.2, using different versions and answers; the score ranks severity, and exploitation sits in separate metrics.
In brief 4 more today
0 of 3 read
Drag a card to start Drag left or right · tap to read more Hover to lift a card · tap to bring it forward Tap again to put it back Tap a card to open it Tap again to close it Tap a headline to read it as a card