Here's How an AI Slowdown Could Actually Be Enforced

Researchers at the University of Toronto released the 'Pacing the Frontier' report, acknowledging that mechanisms to slow AI have not yet been developed. Ideas include independent model audits, compute tracking and GPU modification. Anthropic said Claude performs 26% of its AI research.

• Claude performs 26% of Anthropic's AI research vs 0% in early 2026
• Anthropic spends 6% of its compute budget on AI safety
• Proposed controls include model audits, compute accounting and GPU modification
• Leaders of Anthropic, OpenAI, SpaceXAI and Google DeepMind backed the idea of slowing