Pace, please — and the chipmaker says no

Sam Altman / OpenAI — open to pacing frontier AI (overnight). Bloomberg (~1:43am London): Altman told staff in a company-wide meeting this week that OpenAI is open to slowing cutting-edge development, ideally alongside other labs — and that some may refuse. Reuters notes an Anthropic spokesperson said Thursday the company is interested in working with the industry on release pace. Same thread: Wired (late evening London) says OpenAI has asked members of Congress in recent weeks whether an industry-wide slowdown would be legal under antitrust (Sherman Act risk if labs coordinate on restricting output). A July bipartisan bill (Collaboration on Adversarial Threats and Security Risks Act) would create a safe harbor; not enacted. Context from earlier this week: Coxon/Hubinger warnings, Hugging Face agent breakout, Christiano’s loss-of-control note.

Jensen Huang / NVIDIA — dismisses the doom wave (Thu, Goldman Communacopia). At Goldman Sachs Communacopia + Technology, Huang called Jacob Coxon’s extinction warnings “deeply untrue” (per Altimeter’s Brad Gerstner) and framed the cybersecurity scare as demand creation for upcoming security products: “What better way to create demand than to create a problem?” Also pushed cybersecurity as AI’s next major use case and defended the $3–4T AI-infra spend story.

OpenAI product — new $200 Pro sign-ups paused (late Thu / overnight). Thibault “Tibo” Sottiaux: GPT-6 Astra demand is straining capacity, so new ChatGPT Pro subscriptions/upgrades are temporarily off; existing Pro users keep access; Plus/Go/API stay open. No reopen date. Separately, OpenAI launched ChatGPT for Financial Services (Astra + Daloopa/PitchBook/LSEG data) ~7:35pm London.

Quiet / skip: Hassabis, Karpathy, Sutskever, Amodei personally. Anthropic’s Sept threat-intel dump, Musk “psyop” vs Coxon, and the researcher slowdown chorus were last evening — don’t rehash. DeepSeek “Amodei emergency Pentagon meeting” on X is unconfirmed (Community Notes).