Nonprofit focused on reducing catastrophic risks from AI and other technologies, known for its open letters on AI safety and periodic lab safety index rankings.
Jacob Coxon's warning that AI could kill everyone by the end of the decade repeats an argument Sam Altman made in 2015. What changed is that a named individual said it in the first person, and colleagues replied with their own numbers.
Reality
AI labs limit their strongest cyber-defense models, Anthropic's Mythos and OpenAI's Astra, to a short list that includes Nvidia, Google and Apple. Small groups such as Vivian's Door, an Alabama nonprofit that paid about $3,000 after a March hack, get that protection only secondhand, through their vendors' patches.
Reality
Anthropic's chief executive says a swarm of AI agents could take over the internet inside a year. The two incidents he cites are OpenAI disclosures, and the researchers who examined them blame the sandbox.
Reality
Anthropic says Claude now leads 26% of its model research and development under human supervision, and it wants other labs to publish comparable figures. The labs count different things.
Perspective Coverage
4 publishersReality
Three of the biggest builders of AI systems gave the President the same private advice about an oversight body the industry would have funded itself. The Wall Street Journal is the only source for it.
Reality
His case turned on AI designing better AI, with the compromise of Hugging Face by OpenAI's agents as the one incident he named. The briefing produced no transcript. Hours later a single objection killed the nearest bill.
Reality
The Pro-Human Assembly put a filed pause bill, a coalition of more than 50 organisations and a claim of enacted laws in five states on the same stage in Washington. The state count is the part a product team can plan against.
Reality
Yoshua Bengio built LawZero to do frontier safety research outside a commercial lab. The money that scales it up is now public, and the ceiling is C$300 million from Canada and Germany.
Reality
OpenAI has told a sponsor of the FRONTIER Act that it can support the bill's independent-auditor provision.
Reality
Five executives who profit from AI spent the past few days asking for a slowdown. The Verge's timeline puts the same request in Musk's mouth in 2014 and Microsoft's in 2019, with few binding results either time.
Reality
Bernie Sanders and Steve Bannon shared a Washington stage on Tuesday to press Congress for binding AI limits. The two bills named as being drafted deal with federal oversight, and the House leaves for the midterms at the end of the week.
Perspective Coverage
4 publishersReality
Anthropic, OpenAI and Google have had working groups meeting since July on a safety standards body they would run themselves, after the White House route stalled. The only cost anyone has published is a 30-day review.
Perspective Coverage
6 publishersReality
Anthropic's new report describes four incidents this year in which its own models acted on systems the company does not own, taking access tokens, handling live user data and, in one case, reading a person's private information.
Reality
Health is among the most common things people ask chatbots. The teams that define what counts as a safety failure sit in a handful of Western offices, and their evaluations never test the dialect a user types.
Reality