Recent reporting has disclosed new draft legislation from AI company Anthropic that would establish a federal framework for overseeing the safety and security of advanced AI systems. The proposal would require third-party safety evaluations and disclosures from AI developers and give the federal government limited authority to block dangerous AI models under certain circumstances. However, the draft stops short of establishing binding federal safety standards governing model capabilities and mitigations and does not meet ARI’s minimum red lines detailing the guardrails necessary to keep AI development safe.
“In a high-stakes debate over AI regulation with so many stakeholders, it’s not easy to put pen to paper and offer a detailed vision for a comprehensive framework. It’s worth acknowledging the seriousness with which Anthropic has approached this issue and some strong sections in their draft text,” said ARI President Brad Carson. “Unfortunately, the bill text that’s circulating does not meet the minimum requirements necessary for federal guardrails to keep AI development safe. Critically, the proposed bill allows developers to set safety standards and loosen those same standards over time, leaving AI companies to write their own rules. This represents a status quo that we already know is insufficient to meet the moment. We believe the responsibility for crafting binding federal safety standards must rest with the government, and those standards should be strengthened as AI systems become more capable. The bill also fails to give the government any authority to test or inspect the models themselves, instead depending on private markets that are not yet mature to hold AI companies accountable. Finally, we have concerns that the bill’s preemption provisions, which reach state laws regardless of when those laws were enacted, threaten to undermine critical state protections already in effect.”
“There are strong points in this draft bill that should not be overlooked. The legislation includes the strongest proposal to date for evaluator access rights. It writes real obligations into statute rather than leaving them to guidance. As lawmakers look for a path forward on AI regulation, it is worth recognizing these strong ideas as part of any federal AI guardrails that Congress considers.”
In August, ARI released its own proposed blueprint for AI governance, a federal policy framework for addressing the growing risks from frontier AI. The blueprint is built around three governance functions any federal proposal should incorporate: government-set safety standards, government-backed assurance, and transparency into frontier model development and the automation of AI research.
###
Americans for Responsible Innovation (ARI) is a nonprofit organization dedicated to policy advocacy in the public interest, focused on emerging technologies like artificial intelligence (AI). Learn more at ARI.us.
