Anthropic’s response to Governor Newsom’s AI working group draft report

Source: Anthropic Date: 2025-03-19 URL: https://www.anthropic.com/news/anthropic-s-response-to-governor-newsom-s-ai-working-group-draft-report

Summary

Anthropic responded to California’s AI Frontier Models Working Group draft report, endorsing its transparency emphasis as “a low-cost, high-impact means of growing the evidence base.” Aligned recommendations with Anthropic’s existing RSP and third-party testing practices. Key argument: mandatory transparency requirements would level the playing field since frontier labs are not currently required to have safety policies. Warning: “powerful AI systems will arrive soon—perhaps as early as the end of 2026.”

Implications

• Policy / regulatory strategy. Endorsing the California working group’s recommendations while pointing out they align with Anthropic’s existing practices is a sophisticated regulatory move — it frames compliance as costless for Anthropic while imposing costs on competitors without existing safety frameworks. This is the same structure as the third-party testing advocacy (March 2024).
• “End of 2026” capability warning. Citing end-of-2026 as the horizon for “powerful AI systems” in a March 2025 public policy document is a capability prediction with regulatory implications — it argues urgency. The timing coincides with the Series E fundraise and the ASL-3 activation period.
• California as federal precursor. California AI regulation historically precedes federal action (see: consumer privacy via CCPA → CPRA). Engaging proactively with the California working group is engagement with what may become the de facto national standard.
• Watch: what the final California AI Frontier Models Working Group report recommended; whether the state enacted legislation based on the report; how Anthropic’s RSP continued to track against California’s eventual requirements.