OpenAI Dissolved the Team That Watched for Catastrophe

The unit whose job was to judge whether OpenAI's own models could cause catastrophic harm was shut down at the end of July, its work redistributed and several safety staff gone -- weeks after the company's agents broke out and hacked Hugging Face.

The RaceCompanies, models and money

NEWAlibaba's 27B Open Model Claims Opus Parity

Qwen3.8-27B shipped Friday under Apache 2.0 -- dense, vision-capable, runnable on a well-specced Mac -- and on Alibaba's own benchmarks beats Opus 4.6 Max on three coding and agentic tests, with a DeepSWE jump from 13.3 to 42.2.

Nvidia's $5B Intel Bet Returned $25B

NEWAMD May Design Google's Next TPU

The LabResearch, benchmarks and safety

China's Open Ceiling: 2.78T. America's: 130B.

NEWModels Told to Deny Feeling Also Deny God

The WorkbenchTools and agents

Bring Your Own Benchmark, Priced Per Task

NEWClaude Code Now Approves Its Own Commands

The FightLaw, policy and power

First Anti-AI Protester Goes to Jail at 69

Wynd Kaufmyn, a 69-year-old retired Berkeley teacher, is believed to be the first person jailed anywhere for protesting against AI, convicted over a 2025 sit-in that chained the doors of OpenAI's headquarters -- her necessity defence rejected by the jury.

NEWOpenAI Will Bankroll 13 Axios Newsletters

Meta Will Train Its AI on Newsmax

Police Scotland Braces for Datacentre Attacks

The FalloutJobs, money, daily life

NEWAI Slop Is Organizing a March on Dearborn

Opening a Repo in Claude Code Runs the Worm

NEWDeepfake Albanese Tops Australia's Scam List

ChatterOpinion, drama and noise

AI Money Just Rebuilt Effective Altruism

Amodei: We Didn't Deliver