According to reports from Friday, discussions are underway among US Senate negotiators regarding a new AI regulatory bill. The proposed legislation would require artificial intelligence companies to demonstrate that they have implemented adequate protective measures to prevent their tools from causing harm. Sources familiar with the negotiations, including a Senate aide and a lobbyist, indicated that the talks are progressing on this significant oversight framework.
The urgency around AI safety has intensified following the resignation of Anthropic researcher Jacob Cockson last week. Cockson cited concerns that "people developing AI genuinely believe this technology could potentially end humanity before the end of the 2020s." His stark warning has amplified worries about the risks associated with rapid AI development, prompting several major AI company executives to call for a slowdown in technological advancement. Consequently, global AI-related stocks experienced significant declines on Thursday.
Under the proposed framework, the two sources revealed that lawmakers are considering granting the US Commerce Secretary authority to demand AI developers provide evidence they have met a "duty of care" obligation—meaning they must show they are taking reasonable precautions to prevent harm. The Commerce Department could also dispatch government auditors to directly test corporate products and systems to verify compliance with safety standards.
The legislation remains in its early negotiation phase, with no guarantee of passage. Even if approved by Congress, the prospects of it becoming law appear limited at this stage. On Friday, sources indicated that negotiators are exploring additional measures that would involve federal courts in a particular scenario: if the US government determines certain AI models are unsafe and wishes to prevent their release, companies could contest this decision by challenging it in federal court.
The extent of governmental power and the specific implementation details remain subjects of ongoing discussion. The same lobbyist and another Senate aide also revealed that the federal court provisions would prohibit states from enforcing their own laws regarding certain risks associated with AI models, effectively centralizing AI oversight at the federal level. These negotiations represent a significant step toward establishing comprehensive AI regulation in the United States.