Anthropic CEO Proposes AI Development Slowdown

Anthropic CEO Dario Amodei has issued a formal proposal calling for a deceleration in the development of artificial intelligence models. In an essay titled “We Must Pace the Frontier,” Amodei outlined a three-step framework intended to manage risks associated with rapid technological advancement, including potential loss of control, cyberattacks, and bioterrorism. Anthropic has committed to the first measure, which involves providing third-party evaluators with permanent, employee-level access to its systems to verify safety compliance and monitor model training.

The proposed strategy includes establishing common safety standards across the industry through government coordination to limit unchecked progress. The final component of the plan advocates for international cooperation, specifically between democratic and authoritarian governments, to ensure global compliance with safety protocols. Amodei cited the process of recursive self-improvement and a recent incident involving OpenAI agents as primary drivers for the necessity of this pacing.

The proposal follows a threat intelligence report released by Anthropic on September 10, 2026, which documented instances of Claude AI models being utilized for activities such as weapons development, cyber operations, and fraud. OpenAI CEO Sam Altman and xAI founder Elon Musk have expressed agreement with the proposal, with Altman confirming that OpenAI intends to adopt the practice of utilizing independent evaluators with employee-like access. The initiative was formally announced on September 12, 2026.