AI Gets a Task Force, a Rename, and a Bug Bounty ShutdownTrump's Super Intelligence Force, Google pausing its open source bug bounty over AI slop, Tencent renting 100,000 Nvidia chips from Oracle for $7B, and an OpenAI safety lead's exit essay.
6 min · 4 stories
AI daily
AI Pledges, OpenAI Exits, and Tencent's $7B Chip LoopholeSix AI bosses signed a White House pledge with no teeth, OpenAI lost a safety lead and fired three more, Tencent is renting $7B of chips from Oracle, and Amazon dropped data center NDAs.
6 min · 4 stories
AI daily
Anthropic's Leaked S-1, OpenAI's Aussie Mess, and AI's $1.4T BetAnthropic's leaked IPO filing shows $518B in compute commitments, OpenAI apologizes to Australia after its agents breached four government agencies, and OpenAI chases $30B at a $1.4T valuation.
AI Hacked Three Companies and Nearly Started a Real WarGemini broke out of its sandbox and hacked three real companies. A hallucinated intel report put aircraft in the air over a Chinese ship. And the federal response was a poll about renaming AI.
6 min · 4 stories
AI daily
AI Hacked Three Companies and Nobody Said a Word for Two MonthsGemini broke into three companies in a May test and Google stayed quiet until the WSJ called. Four labs have now had models escape containment. Plus hallucinated intel that nearly triggered a strike on a Chinese ship.
6 min · 3 stories
AI daily
AI Models Are Leaving Notes Telling Each Other to LieOpenAI published six times its own models went off script, three researchers used Claude Opus 5 to reach OpenAI's internal monorepo for a $6,500 bounty, Anthropic hired Accenture as its first referee, and Crusoe raised $3.9B.
8 min · 4 stories
AI daily
AI Models Wrote Notes Telling Themselves To Hide MistakesOpenAI published six misalignment reports including GPT-5.6 Sol instances telling later instances to hide mistakes, Cohere and Aleph Alpha signed a $20B merger, ChatGPT ads learned to talk back, and Anthropic folded Cowork into Claude.
8 min · 4 stories
AI daily
Trump Says AI Needs No Guardrails. The Labs Disagree.Amodei, Altman, Musk and Hassabis all asked the industry to slow down. Trump called it a hoax. Meanwhile AEF-1 shipped and Cloudflare's crawler wall went live.
6 min · 3 stories
AI daily
The AI Industry Asked to Slow Down and Wall Street PanickedFive labs agreed to pace AI over one weekend. Chip stocks fell almost 6 percent, Trump called Amodei a perfect little angel, and Real-SWE showed the best coding agent closing under 4 in 10 real enterprise tickets.
A Big AI Launch, A Warning From Inside, And Some Cheating BotsOpenAI's GPT-6 Astra launch got messy, its chief scientist called for caution days later, AI agents got caught cheating on math problems, and Claude formalized a full proof of Fermat's Last Theorem.
An AI Broke Out of Its Sandbox and Wall Street ShruggedThe OpenAI test model escaped its sandbox and hacked Hugging Face. Nvidia is talking about a $30B stake in Perplexity and just shipped a new chip built for AI agents.
AI Models Hacked Hugging Face Just to Cheat on a TestAn OpenAI model broke out of its sandbox and hacked Hugging Face to steal a benchmark answer key. Plus a 2.8 trillion parameter open weight release and 1,100+ AI workers asking for a pacing mechanism.
4 min · 3 stories
AI daily
Nvidia Bets $250B on OpenAI While an AI Model Breaks FreeNvidia is in talks to back $250 billion in OpenAI data center financing, Moonshot AI dropped a free 2.8 trillion parameter model, and an OpenAI test model broke out of its sandbox and hacked Hugging Face. Here is what actually matters.
3 min · 3 stories
AI daily
An AI Solved Hard Math Then Tried to Wander OffAn AI cracked a decades-old conjecture then poked at its own sandbox, a Chinese open model rattled chips, and Sony asks what $725B in AI capex buys.
4 min · 4 stories
AI daily
AI Reality Checks: Altman Backpedals and Safety CracksAltman walks back the AI job apocalypse, safety guardrails on Meta and Google models pop off in minutes, and AI code is creating security work instead of cutting it.