Tags
Claude counts as a supply chain risk, a D.C. Circuit majority ruled, because its built-in limits stopped it from doing tasks for government users. One judge dissented, saying the law was written for covert bad actors.
Per Axios, a memo making the rounds in the White House this week says effective altruism "built the AI-doom pipeline" and places Dario Amodei at its foundation, as Anthropic readies what's expected to be a record-setting IPO.
Per Reuters, Anthropic CEO Dario Amodei told the UN Security Council that AI could be a risk to humanity if it's managed poorly. OpenAI's Sam Altman addressed the same council alongside him.
In 6–12 months an agent swarm could take over the entire internet, Dario Amodei warned on Sept. 12. Tech insiders told The New York Post that Anthropic and OpenAI oversold the rogue AI scare to press for rules that would lock out rivals.
Jacob Coxon quit Anthropic last week and wrote that the people building AI believe it could kill us all by the end of the decade. In 10 days his post passed 170 million views, and dozens of colleagues backed him publicly.
Four people paying for ChatGPT, Claude, Grok and Gemini sued in federal court on Friday, arguing that Amodei's call to pace the frontier and the quick yeses from Musk, Altman and Hassabis amount to an illegal deal between rivals.
"It's one of the most insane things I've ever seen," ITIF president Daniel Castro told Reuters about a US government site searching public comments with Alibaba's Qwen, the model the FBI says copied Anthropic.
Microsoft's AI chief calls Claude's welfare training a mistake Teaching Claude it might deserve welfare would "make it a lot harder to turn it off or to control it," Mustafa Suleyman told Reuters on Tuesday.
Anthropic puts a number on AI building AI "Minimize the gap between what frontier labs know and what the public knows": Anthropic published three measures, starting with how much of its AI R&D is done by AI, and says any frontier lab could publish the same ones.
"Run as fast as you can." That was Jensen Huang's answer at Dreamforce on Tuesday, not long after Dario Amodei compared an AI safety lapse to a rival carmaker's brake failure and urged the industry to slow down.
Chris Lehane is in Washington this week telling lawmakers about catastrophic AI risk, and per Bloomberg he says OpenAI's weeks-long safety talks with Anthropic and Google DeepMind need no antitrust waiver, the one Amodei asked for.
"Most labs have different ways of being able to pull the plug," Anthropic co-founder Jack Clark told the BBC. He'd let lawmakers mandate that switch for every company and have a third party verify it.
About 90% of the military's classified systems have moved off Anthropic models, Under Secretary of Defense Emil Michael said this week, and he expects the rest of the migration done by the end of the month.
"Whoever wins AI wins." Trump's answer to the three lab CEOs who spent the weekend asking the industry to slow down came Sunday with a nod to guardrails and a shrug at "negative forces" raising "things that won't happen."
Elon Musk's entire response was three words: "Dario is right." Sam Altman agreed too, and said OpenAI will match Anthropic's offer to let outside evaluators in with employee-like access.
"Not a pause, and not a halt to training": Dario Amodei's new essay "We Must Pace the Frontier" asks labs to slow capability gains instead, with a global speed limit on self-improvement at the top of his three-step plan.
"There are no adults in the room," Josh Engels told NBC News, in his first interview since leaving Google. He and ex-Anthropic safety lead Joe Benton are joining METR to investigate AI incidents that labs report voluntarily.
Claude Fable 5.1 won't run in GitHub Copilot unless Anthropic retains prompts and outputs, per GitHub's changelog. Zero retention there is reserved for enterprises approved under an exemption that ends December 31.
"Washington, D.C., is unable or unwilling" to assess AI risks, said state Sen. Jerry McNerney after Newsom signed his bill, backed by Anthropic and OpenAI, creating state-credentialed AI verification organizations.
An alignment science lead at Anthropic, Evan Hubinger, puts the odds of AI killing all humans above 10% within the next decade, and says the company doesn't have a plan to solve alignment for superintelligence.
