Anthropic Urges Governments to Create System for Collective Pause on AI Work

Anthropic PBC called for a mechanism allowing governments and AI developers to jointly decide when to temporarily halt work on advanced artificial intelligence, warning of unforeseen risks as the technology rapidly evolves. The proposal was outlined in a blog post published on June 5, 2026. The startup acknowledged the difficulty of securing global cooperation, noting that nations and companies are locked in a competitive race. Anthropic previously pledged in 2023 to pause dangerous work, but relaxed that commitment earlier this year, citing a policy shift toward prioritizing AI competitiveness and economic growth over safety. The post compared a potential pause protocol to international nuclear regulation, while conceding that AI training runs are easier to conceal than missile silos and incentives to defect are high. Anthropic plans to discuss coordination with policymakers and peers in coming months, even as it continues releasing advanced models and prepares for an initial public offering.