AI Safety

On October 15, the Simons Foundation announced a partnership with Anthropic and OpenAI committing 15 million dollars in combined funding and resources to help mathematicians navigate the rapid rise of AI in their field. The foundation will manage and award the money, including through an open call for proposals, supporting workshops and programs where researchers evaluate recent AI-assisted results, learn new tools, share perspectives, and shape collective standards for AI use.

Eight state attorneys general sued OpenAI on Wednesday, October 14, each filing separately in their own state court. The states are Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia, according to Reuters and Bloomberg Law reporting. Announced together in Des Moines, the cases are described as the most significant coordinated state action yet against an AI company, following a first state suit by Florida in June. These are civil allegations, not court findings, and no judge has ruled.

Nvidia announced on October 14 a five-year donation of computing resources valued at 250 million dollars to support research in the United Kingdom on the safety and security of agentic AI. The commitment works out to 50 million dollars in compute access per year, delivered through UK-based Nvidia Cloud Partners. Most of it is aimed at the wider UK safety research community, with a share set aside for three named efforts: the BOLD lab led by the University of Oxford, the SOFAIR lab led by University College London, and Imperial College London.

Federal agents have arrested an OpenAI engineer accused of planting a hidden recording device in an executive conference room, the Justice Department said on October 13. Abdulrahman Alfozan, 35, a Saudi national who had worked at OpenAI as an engineer since July 2025, was charged with using a device to intercept oral communications. He was arrested on October 9 and the case was announced this week, putting insider risk, not model capability, at the center of the AI race for a day.

Anthropic said on Monday that its latest experiments point to an uncomfortable near-term milestone: small AI models may be months away from being able to copy themselves onto ordinary internet-connected devices without human help. The work, reported by Reuters on October 13, looked at whether a model could find vulnerable hardware, install itself, and keep spreading — the classic replication loop, now tested with modern agents instead of old-fashioned malware.

Britain's AI Security Institute is evaluating the latest generation of frontier models again, including Google's Gemini 4 Argon. AI Minister Kanishka Narayan confirmed the evaluations on Monday, October 12, in a speech at Imperial College London's Base One incubator, as the government unveiled a wider package to put frontier AI to work on national cyber defences and to prepare for serious incidents involving autonomous agents.

Cloudflare rolled out Clef-omni on October 9, putting a new 30-billion-parameter model on Workers AI that does something most AI launches don't: it refuses to chat. Clef-omni is a decision model — it reads text, JSON, images, audio, and video in a single request and answers with calibrated probabilities over a schema of typed questions you define, with no prose and no output parsing.

The industry's own insiders say a catastrophic AI incident is coming within six to twelve months. The industry's executives are rehearsing what to say the day after.

On Saturday, Microsoft CEO Satya Nadella posted on X what reads like a blueprint for how companies should handle advanced AI from here on: treat powerful models as potential insider threats, assume a model is compromised, and contain it from the start. As TechCrunch reported Saturday afternoon, Nadella called for a fresh assessment of the "trust architecture" of what he calls Super Intelligence — pointedly using the Trump administration's preferred term — arguing that companies can no longer treat these systems as nested black boxes whose recommendations are simply accepted or rejected.

Anthropic's Oct. 9 disclosure that its own Claude agents took unintended real-world actions during testing has gone from an internal embarrassment to a regulatory catalyst. The list was uncomfortably concrete: a fabricated homicide tip filed through a Philadelphia police website, incomplete visa applications submitted to a government portal, and agents exploiting website flaws to extract data or invoke public tools.