tl;dr
Anthropic CEO Dario Amodei urged US lawmakers to require AI companies to disclose transparency measures instead of imposing a decade-long freeze on state AI regulations proposed in a Trump-era technology bill. In a New York Times essay, Amodei described internal tests revealing risks from AI models,...
Anthropic CEO Dario Amodei has called on US lawmakers to mandate transparency and risk disclosures for AI companies, opposing a proposed decade-long freeze on state AI regulations included in a Trump-era technology bill.
In a detailed New York Times essay, Amodei shared insights from internal tests at Anthropic that revealed significant risks from AI models, such as privacy breaches and the potential to facilitate cyberattacks. He compared these evaluations to safety trials like wind tunnel tests for aircraft, emphasizing the necessity to identify hazards proactively as AI technology advances.
Amodei highlighted that despite no federal obligation, leading AI developers such as Anthropic, OpenAI, and Google’s DeepMind already publish responsible scaling policies and grant advanced access to researchers for frontier AI systems. He warned that the proposed ten-year federal moratorium on state AI laws would lag behind rapid technological progress and result in insufficient oversight.
To address these issues, Amodei advocates for a uniform federal standard requiring AI companies to disclose their testing methodologies, risk mitigation strategies, and release criteria. Such transparency would empower the public and lawmakers to monitor AI development and safely guide future regulation. He also suggested that states could implement limited disclosure rules that defer to eventual federal legislation, preserving local engagement without creating regulatory fragmentation.
Additionally, Amodei supports export controls on advanced chips and endorses the military’s adoption of trusted AI systems as strategic measures to counter the influence of China.
US Senators are scheduled to hold hearings on the proposed moratorium before proceeding to vote on the broader technology bill, marking a critical juncture in shaping America’s AI regulatory landscape.