Could AI Labs Legally Agree to Slow Down? Antitrust and Safety Debate Heats Up

Following warning signs like the Hugging Face breach, safety researchers and John Schulman urge AI labs to strike pacing agreements on capability growth, though antitrust scholar Erik Hovenkamp notes such deals are typically illegal unless justified by a 'saving humanity' defense.

2026-09-09 ~ 2026-09-10 · 3 related posts

• After the HF attack, researcher calls for possible temporary bans on model capability gains — EthanJPerez · 2026-09-09
• John Schulman: OpenAI and Anthropic Should Draft an AI Pacing Proposal Together — dpaleka · 2026-09-09
• Antitrust scholar: AI labs agreeing to slow innovation is per se illegal — but saving humanity might justify it — Afinetheorem · 2026-09-10