Artificial intelligence news: models, chips, data centers and what changes for people.

Researchers from University College London and UNESCO tested 33 AI models on climate questions and found strong textbook recall but weak reasoning on policy effects, plus a tendency to give in when challenged.

Officials say a run of failures at Microsoft, the FAA, Chase, United, the NSA and Apple were not cyberattacks, yet suspicion of hackers and AI persists, WIRED's Uncanny Valley podcast reported.

Security firms and the FBI link the Go-based tool, which directs AI models to run penetration tests, to attacks on banks, agencies and companies worldwide since late August.
Colette Kress told staff that raises and equity grants will arrive in December, months early, as Nvidia works to keep its workforce steady during a search for a new chief executive.

More than 500 Alphabet employees signed a petition against a policy that ties student access to Search to having some AI features enabled in schools.

Four Baltimore County teens say a canceled Omnilert alert still led to armed officers, handcuffs and searches, and they want a court to bar the system until protocols are fixed.

After outages hit Apple Pay, NASA, the FAA and Microsoft, security figures told TechCrunch that AI-driven vulnerability discovery and aging infrastructure may be factors, though officials see no coordinated attack.

A senior engineering manager role for agentic AI suggests Apple is building its own agent platform, even as the company has urged caution about the risks agents pose to user data.

The AI developer went public on Thursday at $83 per share, with its founders selling stock and big tech partners buying a separate class of shares, while its infrastructure commitments top $1 trillion.
Gemini on macOS can now operate native apps and browse sites in the background for users, with controls to pause, stop or approve actions, though availability is limited to English speakers in the U.S.
Google's Chrome team created a fully controlled copy of the web, with fake shops, social sites and bank dashboards, to train and stress-test Gemini-based browsing agents safely.
As Anthropic lists at $83 per share, Amazon and Google hold stakes worth about $308 billion, and analysts debate whether supplier-funded growth can now sustain itself.
Google's fourth frontier model in three weeks posts leading benchmark scores, arriving just before Anthropic's Nasdaq debut at a valuation of about $1.5 trillion, Decrypt reported.