At a US House hearing, Treasury Secretary Scott Bessent said AI labs should get no liability exemptions and called for more open-source models built in the US

AI developers will face increased legal and financial responsibility for their creations.

Treasury Secretary Scott Bessent stated at a US House hearing that AI labs should not receive liability exemptions for the AI they develop. He argued before the House Financial Services Committee that holding creators responsible for their AI's output is the most effective safety measure. Bessent specifically pushed back against requests from "frontier labs" for such waivers, emphasizing that a "blank check on liability" should not be granted. His remarks came amid growing calls for AI safety measures and a debate about slowing down AI development, with figures like Anthropic's Dario Amodei advocating for a pause. Bessent also called for more open-source AI models to be developed within the United States, positioning this as a way to compete with China and prevent regulatory capture by large AI firms.

Bessent's strong stance against liability exemptions signals a potential shift in how AI development is regulated, moving towards greater accountability for creators. This could increase the operational costs and risk profiles for AI companies, potentially influencing investment strategies and development roadmaps. The call for more US-based open-source models also suggests a strategic push to foster domestic innovation and competitiveness, particularly in response to global rivals like China.

The push for accountability may compel AI labs to invest more heavily in rigorous safety testing and ethical frameworks before releasing new models. Increased competition from domestic open-source alternatives could democratize AI development and challenge the dominance of a few large players. This could ultimately lead to a more diverse and potentially more secure AI ecosystem, but also raises questions about the practicalities of enforcing liability across a wide range of AI applications.