Anthropic’s AI safety proposal falls far short of protecting public health and safety. Congress must go further. We’re facing huge risk from even the current generation of frontier AI models, and that risk will only escalate with further model development. Government’s most basic obligation is to protect public health and safety. In the context of AI, that means Congress must ensure that AI benefits people and doesn’t harm people. Quick, effective, robust regulation is essential. Several provisions in Anthropic’s proposal go beyond what has been proposed in Congress so far, but big picture, this proposal is insufficient to manage the catastrophic AI risks about which Anthropic, other companies, and many experts are warning us. Anthropic’s proposal contains no minimum standards for AI companies’ safety plans or for third party auditors, and the proposal would make it even harder for the federal government or states to nimbly regulate and act to prevent severe harms. If we’re going to have safety standards and auditors — and we must — it’s essential that regulators have flexible authority to ensure they are actually protecting people. The proposal also uses a definition from SB 53 that was intended for risk disclosures — not for broader regulation — and extends that definition to dramatically limit the situations in which the federal government can stop model deployments. SB 53 was the first AI safety regulation in the country. It is a transparency law, and we intended it to be the first step, not the final step, to protect people from catastrophic risk. Misuse of that law’s definition would effectively gut the emergency shutdown provisions in this proposal. The state law preemption in Anthropic’s proposal also goes too far. Given the abuse we’ve seen by the Trump administration — toward Anthropic earlier this year and more recently when Trump and Musk colluded to disable Starlink/Starshield for Ukraine, as part of their corrupt love affair with Putin — it’s perfectly reasonable to be concerned about corrupt government actors using shutdown powers to attack disfavored companies. Yet, given the scale of the risks about which these companies and many others are warning us, as well as the turbulence of our federal politics, tying the hands of regulators in California and other states is not a good idea. There are several stronger policy proposals, around auditing, creation of a regulator, and various other issues, that are under development now. Policymakers should look to these proposals as the basis for action in the coming weeks and months. We have very little time to get this right, and we must act with urgency. We must ensure Congressional action protects public health and safety and doesn’t undermine it.