Sunday, September 13, 2026
The Anthropic CEO's 'pace the frontier' essay lays out three strategies and commits to one unilaterally. Altman says OpenAI will follow, and Musk and Hassabis endorse independent oversight.
Despite a confidential filing, the CEO told Fortune the company will list only when the business and society are ready, and said an AI beyond human control is 'absolutely' possible to build.
The listing could raise as much as $100 billion at a valuation of roughly $2 trillion or more, which would make it the largest IPO in history, with pricing expected before the November midterms.
The Register argues the deal was built to avoid merger review, and that even an unwinding would change little because rival inference chips are ready to fill Groq's place.
CVE-2026-85102 and CVE-2026-85103 allow remote code execution on Security Gateways; fixes shipped September 9 but no public exploit has surfaced yet.
Long skill descriptions, blanket reading requirements, and rigid checkpoints can eat context or make the model stop early, writes OpenAI's Eric Provencher.
On the new StationeryBench, Astra's median progress score is 46 to MolmoAct2's 12, and Cornell's Yoav Artzi calls the result a 'step change in spatial reasoning.'
Vrije Universiteit Amsterdam's Thibault Schrepel expected unguided ChatGPT to hurt students more than help. 'I was wrong,' he writes.