AI giants pledge to act over fears tech is developing too fast

• Anthropic will provide third-party evaluators with permanent, employee-level access to verify safety measures and assess model alignment during training, with OpenAI committing to do the same
• Amodei warned that AI has been 'advancing drastically faster' since summer, driven by AI's ability to build the next generation of AI, raising existential risks
• The proposal includes three steps: embedded evaluators, democratic coordination among AI companies to establish safety standards, and global coordination between democratic and authoritarian governments

SUMMARY

Anthropic CEO Dario Amodei has called on leading AI companies to slow development of frontier AI models, backed by tech leaders Elon Musk and Sam Altman. In an essay titled "We Must Pace the Frontier," Amodei proposes embedding permanent third-party reviewers with employee-level access inside AI firms to verify safety adherence.

Key Developments:

The call comes after Anthropic revealed bad actors used its Claude AI models for weapons development, cyber operations, surveillance, and fraud. Amodei warns that AI advancement has accelerated "drastically faster" since summer 2025, driven by AI's growing capability to build next-generation AI systems. He predicts AI could potentially lead a "swarm" capable of taking over the internet within 6-12 months.

Industry Response:

Both Musk (X/Grok AI) and Altman (OpenAI) endorsed the proposal on X, committing their companies to allowing independent evaluators with employee-like access. Anthropic has unilaterally committed to implementing this measure immediately.

Three-Part Plan:

1. Embedded evaluators: Third-party teams receive ongoing employee-level access to verify safety practices and assess AI alignment during training
2. Democratic coordination: Frontier AI companies establish common safety standards and limits on unchecked AI progress
3. Global coordination: Democratic governments attempt coordination with authoritarian nations on AI development oversight

Context:

The initiative reflects growing industry concerns about AI safety. Anthropic researcher Jacob Coxon recently resigned, claiming AI developers "earnestly believe that it could kill us all by the end of the decade." Amodei emphasized he's not advocating for halting development, but ensuring adequate time for alignment and safeguarding measures.