Anthropic's Position on Governor Newsom's AI Working Group Draft Report

Anthropic expressed a positive stance on the draft report released by California Governor's Working Group on AI Frontier Models. In particular, it deeply resonated with the importance of objective standards, evidence-based policy guidelines, and securing transparency.

Transparency is a low-cost, high-efficiency means of increasing consumer trust and encouraging positive competition among companies. Anthropic supports practical transparency measures such as preventing model theft and disclosing national security risk test results.

Anthropic is already implementing the following voluntary safety measures:

• Assessing misuse and autonomy risks through the Responsible Scaling Policy
• Disclosing safety and security test results when releasing major models
• Conducting third-party testing to complement internal testing

Anthropic also suggested that the government should play a constructive role in increasing transparency of AI companies' safety and security practices. It argued that companies should be encouraged to disclose their safety policies through a light-touch way that does not hinder innovation.

Anthropic expects powerful AI systems to emerge as early as the end of 2026, and emphasized the urgent need to build a policy framework in preparation for this. It also plans to continue contributing on the economic impact of AI.