At a US House hearing, Treasury Secretary Scott Bessent says AI labs should get no liability exemptions and calls for more open-source models built in the US
Matt Bracken /FedScoop:
Context & Ripple Effects
The hearing lands as Senate negotiators weigh a duty-of-care framework for AI companies that could allow the government to block models considered unsafe. OpenAI had already rejected a proposed Illinois liability shield while backing stronger safety legislation, creating a concrete precedent for labs accepting obligations rather than blanket protection.
Bessent’s position also joins AI oversight to competition with China: Treasury had said it would examine whether Chinese models were distilled from US systems and consider sanctions. Calling for US-built open-source models puts model availability alongside accountability in that policy frame.
First-order effects
- AI labs have no Treasury endorsement for liability exemptions, strengthening lawmakers’ leverage to seek enforceable safety duties rather than broad carve-outs.
- US open-source model developers gain high-level political support as Bessent frames domestic model availability as part of competition with China.
Second-order effects
- Senate negotiators considering the proposed duty-of-care bill face a sharper divide between rules governing harmful model releases and claims that industry needs legal protection to develop competitively.
- Labs seeking government legitimacy must reconcile safety commitments with an agenda that favors wider domestic access to open-source models.
Third-order effects
- If this pairing holds, US AI policy will increasingly treat liability rules and model access as linked instruments of industrial competition, rather than separate safety and innovation debates.
- The policy advantage may shift toward labs able to demonstrate both accountability to lawmakers and alignment with US strategic objectives.
The trend: US AI governance is moving toward a state-compatible model in which legal accountability and domestically controlled model access are both tests of strategic legitimacy.