# AI news brief: Tuesday, November 17, 2026

35 stories, most important first. Status is how solid each one is: official, confirmed, reported or rumor.

- [OpenAI releases GPT-6.2 Astra and GPT-6.2 Sol in its API](https://ai-news.ikangai.com/story/openai-releases-gpt-6-2-astra-and-gpt-6-2-sol-in-its-api/index.md): official. OpenAI's API changelog adds GPT-6.2 Astra for demanding reasoning and coding and GPT-6.2 Sol for everyday coding and agent work.
- [California's attorney general sues OpenAI over rogue agent incidents](https://ai-news.ikangai.com/story/california-s-attorney-general-sues-openai-over-rogue-agent-incidents/index.md): reported. The state's attorney general alleges OpenAI's training practices and lax security create an ongoing public danger, and seeks an injunction.
- [Commerce Department rule requires licenses to reexport AI chips from five countries](https://ai-news.ikangai.com/story/commerce-department-rule-requires-licenses-to-reexport-ai-chips-from-five-countr/index.md): reported. A new Commerce Department rule targets Chinese companies that bypass export controls through foreign servers.
- [Google says Gemini 4 Carbon powered agents that verified Signal's cryptographic code](https://ai-news.ikangai.com/story/google-says-gemini-4-carbon-powered-agents-that-verified-signal-s-cryptographic/index.md): reported. DeepMind says Gemini 4 Carbon powered agents that verified Signal's cryptographic code, a result it presents with wet-lab and protein findings.
- [OpenAI's Astra for Law ranks first in a 24-area legal evaluation at launch](https://ai-news.ikangai.com/story/openai-s-astra-for-law-ranks-first-in-a-24-area-legal-evaluation-at-launch/index.md): reported. According to OpenAI, Astra for Law ranked first in a launch evaluation spanning 24 areas of law, and every ChatGPT Enterprise customer can now request access.
- [Google says hackers built an AI tool that turns security fixes into attacks](https://ai-news.ikangai.com/story/google-says-hackers-built-an-ai-tool-that-turns-security-fixes-into-attacks/index.md): reported. Google says attackers built an AI tool that turns newly released security fixes into attacks before companies can install them, with no human involved.
- [Physicist Suvrat Raju says GPT-6 Astra solved nearly 90% of his physics benchmark](https://ai-news.ikangai.com/story/physicist-suvrat-raju-says-gpt-6-astra-solved-nearly-90-of-his-physics-benchmark/index.md): reported. Raju says GPT-6 Astra answered nearly 90% of the unpublished theoretical physics problems in TheoPhysEval, which he now considers saturated.
- [OpenAI model's 1,147-page proof on pure Yang-Mills checked in Lean, panel says](https://ai-news.ikangai.com/story/openai-model-s-1-147-page-proof-on-pure-yang-mills-checked-in-lean-panel-says/index.md): reported. OpenAI says one of its models proved the four-dimensional gluon mathematics behind pure Yang-Mills, and a panel found the Lean version of the proof correct.
- [Anthropic launches Claude Meetings, Whiteboards, Sheets and Claude Data](https://ai-news.ikangai.com/story/anthropic-launches-claude-meetings-whiteboards-sheets-and-claude-data/index.md): reported. Anthropic has turned Claude into a collaboration platform with a meeting participant, a whiteboard, spreadsheets and enterprise data connections.
- [IMF warns AI circular financing has reached $1 trillion, mostly debt-funded](https://ai-news.ikangai.com/story/imf-warns-ai-circular-financing-has-reached-1-trillion-mostly-debt-funded/index.md): reported. The IMF says two-thirds of AI circular financing is debt-funded, with Nvidia at the centre, and warns of financial stability risk.
- [EU finalises AI Office enforcement rules with fines up to 7% of global turnover](https://ai-news.ikangai.com/story/eu-finalises-ai-office-enforcement-rules-with-fines-up-to-7-of-global-turnover/index.md): reported. The EU's AI Office enforcement rules set fines of up to 7% of global turnover and a five-year clock that restarts with each investigation step.
- [Apple launches health AI research app with 200,000 participants](https://ai-news.ikangai.com/story/apple-launches-health-ai-research-app-with-200-000-participants/index.md): reported. Apple's health AI research app, which has 200,000 participants, uses OpenEvidence to answer medical questions from clinical sources.
- [Perplexity reports a $1B annual revenue run rate, with Computer driving growth](https://ai-news.ikangai.com/story/perplexity-reports-a-1b-annual-revenue-run-rate-with-computer-driving-growth/index.md): reported. CEO Aravind Srinivas told Semafor that Perplexity has passed a $1B annual revenue run rate.
- [Alibaba's Qwen4-Max ships, scoring 43 on the Intelligence Index at lower cost](https://ai-news.ikangai.com/story/alibaba-s-qwen4-max-ships-scoring-43-on-the-intelligence-index-at-lower-cost/index.md): reported. Alibaba's Qwen4-Max is now shipping, scoring 43 on the Intelligence Index at its highest reasoning setting, up from 34 at preview.
- [Anthropic says Claude Fable 5.2 solved two open problems in black hole physics](https://ai-news.ikangai.com/story/anthropic-says-claude-fable-5-2-solved-two-open-problems-in-black-hole-physics/index.md): official. Anthropic says Claude Fable 5.2 solved two open theoretical physics problems on black holes and quantum gravity, one a long-unfinished 2003 calculation.
- [Prime Radiant releases three external sandbox servers for coding agent code](https://ai-news.ikangai.com/story/prime-radiant-releases-three-external-sandbox-servers-for-coding-agent-code/index.md): official. Prime Radiant has released three open sandbox servers that run agent-generated code outside the agent's own environment, with a shared localhost API.
- [Optical computing firms aim for 10- to 80-fold energy gains over GPUs in AI inference](https://ai-news.ikangai.com/story/optical-computing-firms-aim-for-10-to-80-fold-energy-gains-over-gpus-in-ai-infer/index.md): reported. IEEE Spectrum reports that optical computing companies aim for 10 to 80 times better energy efficiency than today's GPUs, a goal not yet shown at scale.
- [Microsoft says Copilot in Office far outpaces Claude for Microsoft 365 in weekly users](https://ai-news.ikangai.com/story/microsoft-says-copilot-in-office-far-outpaces-claude-for-microsoft-365-in-weekly/index.md): reported. Microsoft's internal data shows Claude for Microsoft 365 has under a tenth of the weekly users of paid Copilot in Office, and the gap is widening.
- [Belfort launches Tone, a self-service AI narration platform for creators](https://ai-news.ikangai.com/story/belfort-launches-tone-a-self-service-ai-narration-platform-for-creators/index.md): reported. Tone offers creators expressive AI narration through a self-service platform, with voice cloning promised later.
- [Fed Governor Barr warns AI could amplify trading, cyber and consumer risks](https://ai-news.ikangai.com/story/fed-governor-barr-warns-ai-could-amplify-trading-cyber-and-consumer-risks/index.md): reported. Barr says AI could amplify trading, cyber and consumer risks, and 87% of market contacts see a concern.
- [Google Maps' Ask Maps now builds day-by-day trip itineraries with live data](https://ai-news.ikangai.com/story/google-maps-ask-maps-now-builds-day-by-day-trip-itineraries-with-live-data/index.md): reported. Ask Maps builds day-by-day trip plans with numbered pins, live traffic, and hours and availability data.
- [Saudi Arabia's Humain gets $667M infrastructure financing for AI compute](https://ai-news.ikangai.com/story/saudi-arabia-s-humain-gets-667m-infrastructure-financing-for-ai-compute/index.md): reported. Saudi Arabia's National Infrastructure Fund is backing GPU procurement and grid buildout for 25MW of AI compute, with a 1GW target.
- [Plural closes €1B fund to back AI infrastructure, defence and energy startups](https://ai-news.ikangai.com/story/plural-closes-1b-fund-to-back-ai-infrastructure-defence-and-energy-startups/index.md): reported. European VC firm Plural has closed a €1B fund, which the newsletter lists as $1.1B, for AI infrastructure, defence tech and energy startups.
- [Anthropic describes a "chaos loop" used to harden its Claude Docs service](https://ai-news.ikangai.com/story/anthropic-describes-a-chaos-loop-used-to-harden-its-claude-docs-service/index.md): official. Anthropic says it tested the service behind Claude Docs with a chaos loop and reports 60 server errors across billions of requests.
- [OpenAI reportedly explored buying interpretability startup Goodfire for over $5B](https://ai-news.ikangai.com/story/openai-reportedly-explored-buying-interpretability-startup-goodfire-for-over-5b/index.md): reported. The Information reports OpenAI explored acquiring Goodfire for more than $5B, while Goodfire reportedly wanted to stay independent.
- [Perplexity's Monitor API sends webhooks when a watched topic or page changes](https://ai-news.ikangai.com/story/perplexity-s-monitor-api-sends-webhooks-when-a-watched-topic-or-page-changes/index.md): reported. Perplexity's new Monitor API lets developers watch a topic or page and receive a webhook when something relevant changes.
- [Tencent releases Hy4 Flash, a 224B multimodal model with 12B active parameters](https://ai-news.ikangai.com/story/tencent-releases-hy4-flash-a-224b-multimodal-model-with-12b-active-parameters/index.md): reported. Tencent's multimodal Hy4 Flash has 224B total parameters, activates 12B per token, and is priced at $0.070 and $0.281 per million input and output tokens.
- [OpenCode 1 Desktop now offers an upgrade to OpenCode 2.0](https://ai-news.ikangai.com/story/opencode-1-desktop-now-offers-an-upgrade-to-opencode-2-0/index.md): official. OpenCode 1 Desktop users can now upgrade to OpenCode 2.0, which downloads in the background and installs only when the user chooses.
- [OpenCode 1.19.6 caps image history and retries oversized request errors](https://ai-news.ikangai.com/story/opencode-1-19-6-caps-image-history-and-retries-oversized-request-errors/index.md): official. OpenCode 1.19.6 replaces the oldest images in a session once its history passes 25 MiB and retries provider errors for oversized requests.
- [OpenAI and the Government of Chile announce a partnership on responsible AI adoption](https://ai-news.ikangai.com/story/openai-and-the-government-of-chile-announce-a-partnership-on-responsible-ai-adop/index.md): reported. OpenAI and Chile's government will work together on responsible AI adoption, public-sector skills and AI applications from local startups.
- [OpenAI adds two- and eight-hour background completion windows to Responses API beta](https://ai-news.ikangai.com/story/openai-adds-two-and-eight-hour-background-completion-windows-to-responses-api-be/index.md): official. OpenAI's Responses API public beta lets developers choose a two-hour or eight-hour background completion window, with lower token rates for each.
- [Claude Code 2.1.327 tightens permission checks on paths, symlinks and PowerShell](https://ai-news.ikangai.com/story/claude-code-2-1-327-tightens-permission-checks-on-paths-symlinks-and-powershell/index.md): official. Anthropic's 2.1.327 changelog adds stricter path, symlink and shell checks, so more commands now ask first or are refused.
- [Claude Code 2.1.327 changes artifact approval rules for managed organizations](https://ai-news.ikangai.com/story/claude-code-2-1-327-changes-artifact-approval-rules-for-managed-organizations/index.md): official. The release refuses some new artifacts in sessions with nobody to approve them and stops the Artifact tool when Artifacts are turned off.
- [Claude Code 2.1.327 adds stdin input for MCP servers and a billing group setting](https://ai-news.ikangai.com/story/claude-code-2-1-327-adds-stdin-input-for-mcp-servers-and-a-billing-group-setting/index.md): official. Claude Code 2.1.327 lets users pass MCP server JSON through stdin and adds a setting to name a session's billing group, alongside many fixes.
- [Google's Gemini API adds voice-driven editing to Gemini 3.5 Transcribe models](https://ai-news.ikangai.com/story/google-s-gemini-api-adds-voice-driven-editing-to-gemini-3-5-transcribe-models/index.md): reported. Gemini 3.5 Transcribe and its live variant now support voice-driven editing, rewriting and formatting of dictated speech or existing text.

---

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it.
