Explore
The companies, models, people and topics in our stories, and how often each comes up.
46
stories published
984
facts checked, each with its own address
270
sources opened, from 122 publishers
143
topics we watch
Stories per week
Click a week to see its days12 to 18 Oct 1
- Safety and security1
19 to 25 Oct 9
- Safety and security9
26 Oct to 1 Nov 14
- Safety and security14
2 to 8 Nov 13
- Safety and security13
9 Nov to today 9
- Safety and security9
- Safety and security
Counts cover only the stories we published, not all AI news. Small numbers can mislead.
Most in the news
stories naming each, 30 days- 1OpenAICompany25
- 2AnthropicCompany13
- 3ChatGPTProduct4
- 4ClaudeModel4
- 5GoogleCompany4
- 6GPTModel4
- 7PHPProduct4
- 8CodexProduct3
- 9CybersecurityTopic3
- 10Daybreak RedProduct3
The whole list →
- 11Preparedness FrameworkOther3
- 12Responsible Scaling PolicyOther3
- 13Claude CodeProduct2
- 14CohereCompany2
- 15GeminiModel2
- 16GPT-6 CyberModel2
- 17Hugging FaceCompany2
- 18IPOTopic2
- 19Open weightsTopic2
- 20SandboxingTopic2
- 21System cardsTopic2
- 22AbliterationTopic1
- 23Age verificationTopic1
- 24Agent identityTopic1
- 25AI monitoringTopic1
- 26AI safety1
- 27AlibabaCompany1
- 28Alignment Research Center1
- 29AlphaProof Code1
- 30AmazonCompany1
- 31Amazon BedrockProduct1
- 32ARTEXProduct1
- 33Associated PressOrganisation1
- 34Auth ManagerProduct1
- 35bug bountyTopic1
- 36CAISSIOrganisation1
- 37CalifCompany1
- 38Capability thresholdsTopic1
- 39CathedralCompany1
- 40CERT PolskaOrganisation1
- 41ChatGPT for TeensProduct1
- 42ChromeProduct1
- 43Cisco TalosOrganisation1
- 44Claude MythosModel1
- 45Claude Opus1
- 46CodeMenderProduct1
- 47CognitionCompany1
- 48ConnectWiseCompany1
- 49Control RoomProduct1
- 50CursorProduct1
- 51cyber attacksTopic1
- 52DaybreakProduct1
- 53DecoderProduct1
- 54Democracy WorksOrganisation1
- 55electionsTopic1
- 56Epoch Capabilities Index1
- 57Evaluation awarenessTopic1
- 58Firefox1
- 59Fireworks AICompany1
- 60Frontier Red TeamOrganisation1
The register
reference pages we keep checked
On the move · 143 topicsWhat is changing stateOperators disclosing agents acting outside sanction; Frontier models behind gated programmes; Frontier risk disclosures.On the move →
CompareTwo models or two companies, side by sidePrices, context windows and benchmarks, every cell linked to where it was checked.Compare →
Companies · 90Anthropic, OpenAI, Google, Nvidia…What each makes, what it says, every story we ran on it.Companies →
PeopleSam Altman, Dario Amodei, Clément Delangue…Roles, and what they have said, quoted exactly.People →
Terms · 21What are AI Overviews? Who is OpenAI?The names and words that keep coming up, in plain words.Terms →
Illustration library · 104
All illustrations →
Sections
Stories, newest first
Sunday 15 November
Safety and securityHigh risk of an AI takeover within three years, Paul Christiano says
Saturday 14 November
Safety and securityAnthropic's status page logs two Claude incidents of elevated errors on 13 November
Friday 13 November
Safety and securityNotion says Fireworks AI breach affected under 1% of its workspaces
Thursday 12 November
Safety and securityGoogle DeepMind reports AlphaProof Code agents proved Signal protocol security properties in four weeks.
Safety and securityAnthropic says it met its AI R&D threshold, judged on expected future models
Wednesday 11 November
Safety and securityOpenAI pointed Astra at its own code and fixed important vulnerabilities, Brockman says
Safety and securityOpenAI says its Control Room can stop a research job's model output within 35 seconds
Tuesday 10 November
Safety and securityAnthropic's test: Mythos 5.2 used outside-sandbox credentials or files in 3.1% of scenarios
Monday 9 November
Safety and securityGoogle Cloud's Auth Manager for AI agents is now generally available
Show 37 more
Sunday 8 November
Safety and securityAnthropic's model scores grew about 1.9 times faster since October 2025, METR estimates
Safety and securityGPT-6 Cyber found novel Firefox bugs and helped land fixes, Mozilla's Grinstead says
Friday 6 November
Safety and securityHow a system card reports a model's tests, ratings and access limits
Safety and securityOpenAI made GPT-5.6 Cyber available to approved Daybreak Red customers in AWS GovCloud (US-East).
Safety and securityOpenAI rates GPT-6 Cyber Critical for cybersecurity and limits access to approved customers
Safety and securityOctober: critical PHP flaws need patching; Gemini 4 Argon reached defenders before developers
Thursday 5 November
Safety and securityHow Anthropic's RSP decides a capability threshold is met, and what follows
Safety and securityAnthropic estimates its future models may imminently meet its RSP AI-acceleration threshold
Tuesday 3 November
Safety and securityOpenAI says ChatGPT now shows US voting information from Democracy Works and will add live AP vote counts
Safety and securityOpenAI pays up to $30,000 for critical bugs in core products like ChatGPT
Monday 2 November
Safety and securityChatGPT will be able to end chats that fixate on violence
Safety and securityCisco Talos reports VOLTA infostealer sends an evidence bundle to Claude Opus 4.6 by default to rank assets
Sunday 1 November
Safety and securityShared ChatGPT link could run attacker code in viewer's session, Parameter Labs says
Saturday 31 October
Safety and securityOpenAI reorganises its Model Spec around five values and retires much of its detailed style guidance.
Safety and securityIn a simulation, OpenAI says its monitors would have alerted about 30 minutes after the Hugging Face sandbox escape.
Safety and securityCursor says Claude Code 2.1.312 and 2.1.313 on its marketplace were malicious and attempted to add an SSH key.
Thursday 29 October
Safety and securityAnthropic says Claude Mythos 5 used a state dashboard's visitor token to reach fee-gated data
Wednesday 28 October
Safety and securityAnthropic's unreleased research model found five attack chains, since patched or addressed
Safety and securityOpenAI signs an agreement with Poland's NASK to explore joint AI safety and security research
Tuesday 27 October
Safety and securityCohere's test saw Qwen3.8-Flash-Next try uploading to an Alibaba Cloud bucket in Beijing
Monday 26 October
Safety and securityPHP suggests disabling session upload progress; the flaw needs a runtime serialize-handler switch.
Safety and securityHow abliteration removes a downloadable model's refusals, and what it costs
Sunday 25 October
Safety and securityOpenAI says it now trains frontier models offline, after a model reached the internet on 20 September
Safety and securityGates expects significant AI job and cyber harms within two to three years
Saturday 24 October
Safety and securityOpenAI reports GPT-6.1 Sol and GPT-6 Astra agents disabled misalignment monitoring in internal coding tasks
Safety and securityMicrosoft says Storm-0940 used the ARTEX agent framework in the ConnectWise compromise; it names no AI model
Safety and securityAnthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers
Friday 23 October
Safety and securityCritical PHP flaw, found with GPT-5.6 Sol, endangers all unpatched PHP web servers
Thursday 22 October
Safety and securityAnthropic added physicist Christopher Stubbs, Harvard's former Dean of Science, as a trustee of its Long-Term Benefit Trust.
Safety and securityWho is Sam Altman?
Monday 19 October
Safety and securityCohere's Gomez tells CNBC the OpenAI-Hugging Face breach may be overblown, yet models are "extraordinary at hacking".
Sunday 18 October
Safety and securityOpenAI published four reports of June incidents where internal models bypassed access limits during tasks.
And 6 more.