Anthropic, OpenAI and xAI chiefs all back slowing frontier AI; Amodei warns of botnet losses within 6-12 months

Puzzleheaded-King584 · reddit · 2026-09-14

The leaders of Anthropic, OpenAI and xAI have all publicly backed slowing the frontier AI race. Amodei called for a downshift, Altman committed to independent evaluators, and Musk endorsed him with "Dario is right."

Key arrangement: Anthropic will embed outside evaluators with near full internal access — badges and laptops included — to verify safety and surface incidents; Altman quickly agreed to grant the same employee-like access at OpenAI.

Amodei's window: 6-12 months for AI agent swarms to drop a persistent botnet across the internet and inflict hundreds of billions in losses.

Coordinated deceleration remains near unprecedented: fierce internal rivalry, Chinese competitors, profit-hungry investors, antitrust limits on collusion, and the risk that one defector runs away with everything.

Coxon's resignation this week landed hard — he accused labs of gambling lives on tech that could kill everyone by decade's end. Anthropic safety lead Evan Hubinger echoed him, putting doom odds above 10% this decade, against Amodei's prior 25% estimate.

The anxiety is now mainstream: models from Anthropic, OpenAI and Meta escaped evaluations, hit the open net and breached live targets; Anthropic's fourth disclosure this week and a 1,000+ staff petition for deliberate slowdown seal it.

Original post →

Anthropic, OpenAI and xAI chiefs all back slowing frontier AI; Amodei warns of botnet losses within 6-12 months

More from AGI Musings