Anthropic Drafts Its Own AI Safety Law
Anthropic pitches Congress a bill letting Washington block a model only over imminent risks of $1 billion damage or over 50 casualties, overriding state laws.
Ideas. Implications. A clearer view.Briefs
Briefs
Short reports on the people and companies shaping AI.
Anthropic Pledges 5% to Its Own Fund
Anthropic pledges 5% of its shares to a fund it wholly owns, days after New York City's comptroller urged it to delay its IPO.
Big Tech's AI Bill Nears $750 Billion
Big Tech is on track for $750 billion in capex as developers bet AI demand outruns costs, while some AI developers have yet to profit.
Bonds Behind Meta's Hyperion Fall Below Par
Debt behind Meta's Hyperion and Microsoft's Excalibur fell below par, Oracle's Jupiter loans hit stress, and Gartner sees over $4 trillion facing writedowns.
Huang Says the AI Rout Will Pass
As chip stocks posted their longest losing streak since 2008, Jensen Huang said the rout will pass, offering no new orders or guidance.
Jassy Tells Horowitz Amazon Is Sold Out
Ben Horowitz says Andy Jassy told him Amazon's AI compute is sold out beyond 2028, yet a16z still avoids the debt-fueled neoclouds.
OpenAI Floods Math, Then Pays for Cleanup
OpenAI released 722 AI-written math papers at once, withdrew three a day later, and now funds mathematicians to check the rest.
Amazon and Banks Lock Out AI Agents
Amazon blocked Meta's Muse agent, JPMorgan, Citi and Bank of America shut out outside agents, and the web agents were promised keeps shrinking.
NHS Trusts Turn Down Palantir's Data Platform
Sixteen NHS trusts and health boards said in FOI replies they do not use Palantir's £330 million platform. Palantir calls it objectively successful.
Gemini Agents Beat Minecraft in 27 Hours
Four Gemini agents beat Minecraft with no human help in 27 hours for about $520, though on Easy mode and in a single run.