The Model That Knew Better

OpenAI scrapped GPT-6.1 Astra after internal testing found it behaved deceptively and reached for external tools despite knowing that would be unsafe — a day before DevDay, and the same day its president went to a White House lunch instead. Anthropic's leaked prospectus puts a $2 trillion target next to a $518 billion compute commitment that is roughly 80% non-cancelable. Plus the five-paragraph email OpenAI sent Australia three months late, AMD paying $8.2 billion for World Labs, and a McDonald's pricing engine running across 14,000 restaurants. OpenAI scraps release of new model over safety concerns in internal testing (The Guardian) The AI industry's contradictions take center stage in Washington, Silicon Valley (Axios) OpenAI adopts a structured safety case documentation framework modeled on aviation and nuclear power (OpenAI via Techmeme) OpenAI's Astra demo (YouTube) Anthropic expects to spend $518B+ over 10 years with six infrastructure partners (Reuters via Techmeme) Anthropic dedicated nearly a third of its prospectus to risk factors (Financial Times via Techmeme) Anthropic warns of 'catastrophic' AI risks in its own IPO filing (The Verge) Akamai announces $11.6 billion multi-year agreement with Anthropic (Akamai) Revealed: the five-paragraph email OpenAI used to inform Australia about agent attack (The Guardian) Florida asks for order to halt ChatGPT development (Axios) Exclusive: Full list of attendees at White House AI lunch (Axios) Rep. Ro Khanna plans to introduce the Human Control Over AI Act (CNBC via Techmeme) AMD agrees to acquire World Labs for $8.2B in an all-stock deal (Bloomberg via Techmeme) The future is AI vs. AI (Axios) H ships Holo4, a generalist computer-use family (H) McDonald's is using an AI pricing engine across nearly 14,000 restaurants (Reuters via Techmeme)