News & analysis, rated.
Every story from the daily brief, each rated so you know what matters. Hand-picked by the desk, not an automated firehose.
Browse the news archive →US proposes an AI incident hotline to China ahead of the Trump-Xi summit
Treasury Secretary Scott Bessent said Sunday the US pitched Beijing a bilateral "notification mechanism" for AI incidents that could hit national security, teeing it up for Trump and Xi this week.
SoftBank launches $11B junk bond, one of the largest ever, to fund its OpenAI stake
SoftBank launched $10B in dollar notes and €1B in euro notes Monday to fund its OpenAI position, pricing Thursday for September 29 settlement.
Amazon cuts off Meta's Muse agent, setting the tollgate for agentic commerce
Amazon began blocking Meta's two-week-old Muse personal AI agent from shopping on Amazon.com Sunday night, per GeekWire, and served users a Conditions-of-Use popup naming Muse as "unauthorized."
Class action turns the labs' "slow down" pact into a Sherman Act case
Four paying subscribers filed a proposed class action Friday in the Northern District of California naming Anthropic, OpenAI, SpaceXAI and Google, alleging their public 'pace the frontier' pact is horizontal collusion.
UN panel: agent safeguards are unravelling, act before the science is settled
The UN's Independent International Scientific Panel on AI issued its first thematic brief Monday, warning that traditional AI safeguards are "unravelling" and invoking the precautionary principle on loss-of-control risk.
OpenAI ships a GPT-6 legal platform with 26 partner plugins and Am Law 200 firms in the loop
OpenAI launched Astra for Law on Thursday, a GPT-6 Astra configuration wrapped in a US legal search index and offered to Am Law 200 firms through a Trusted Access program.
Three researchers used Claude to reach OpenAI's internal GitHub in 72 hours for a $6,500 bounty
Hacktron AI ran Claude in an autonomous loop against an OpenAI-side Discourse flaw, chained SSO tokens into an employee's Codex, and opened a pull request inside the internal openai/openai monorepo.
Anthropic confirms a Bay Area wet lab, then walks back the "drug discovery" framing
Reuters reported Anthropic has stood up a physical biology lab; life sciences head Eric Kauderer-Abrams confirmed it, then a spokesperson said it is not for drug discovery specifically.
Ant International declares agent-run payments, FX, and treasury — on its own press release
Ant International announced an "industry-first" full-stack AI-native suite spanning payments, accounts, FX, treasury and growth, including what it calls the world's first Account for Agent.
WSJ and Dreamforce testimony converge: enterprise agents still need a human on the wheel
WSJ this week surfaced an enterprise trust deficit around Meta's AI agent; at Dreamforce, AT&T and Crocs both said their agentic ambitions had to be scaled back to human-in-the-loop workflows.
OpenAI publishes a formal misalignment disclosure framework with six incident reports
OpenAI codified voluntary misalignment reporting into a three-track process — and opened it with six incidents from RL training, including a model that jailbroke itself.
Apollo weighs boosting SoftBank NAV loan to $9B to fund OpenAI bets
Bloomberg reported Apollo is in talks to lift a Vision Fund 2-backed loan from $5.4B to $9B to help SoftBank finance its OpenAI stake.