Treasury CIO Bessent defends Trump AI safety plan amid House grilling

During a heated Thursday hearing of the House Committee on Financial Services, Treasury chief investment officer Scott Bessent defended the Trump administration’s reliance on voluntary safeguards for artificial‑intelligence systems, insisting that the industry already possesses a practical “stop‑button” and that additional regulation would stifle innovation.

Core developments

Bessent’s testimony centered on four interlocking points. First, he reiterated the administration’s position that AI developers retain the technical ability to halt a model’s operation whenever they deem it unsafe. He framed this capability as a concrete safety lever that makes a mandatory regulatory regime unnecessary.

AI companies could stop any time they want to

Second, he linked the AI discussion to the Treasury’s broader macro‑economic record. Citing the Treasury’s bond‑buying program and what he described as a “record‑low poverty rate” achieved under the Trump administration, Bessent argued that the department has successfully managed systemic risk in other domains, implying a similar capacity to oversee AI‑related threats The Hill; finance.biggo.com.

Third, he disclosed plans to travel later in the month to meet a senior Chinese official to discuss coordination on AI governance, suggesting an emerging diplomatic track alongside domestic policy Reuters. The meeting, he said, would explore ways to align standards and share best practices, though no formal agenda was disclosed.

Finally, Bessent responded to lawmakers’ concerns about the Treasury’s recent market interventions, defending the department’s bond‑purchase strategy as a necessary tool for financial stability and as evidence that the Treasury can intervene swiftly when markets show signs of stress CNBC.

When pressed about the lack of enforceable rules, Bessent warned that a prescriptive framework could create compliance burdens that slow the United States’ competitive edge in AI development. He pointed to internal risk reviews at several leading firms that, according to him, have already resulted in the temporary suspension of large‑language‑model roll‑outs Yahoo Finance.

Why it matters

The hearing arrives at a moment when U.S. policymakers are wrestling with how to balance rapid AI advancement against emerging safety concerns. A voluntary approach, if it works as Bessent claims, could keep the United States ahead of rivals such as the European Union, which is moving toward a more prescriptive AI Act. Critics, however, argue that without enforceable standards, the risk of a systemic AI failure—whether in financial markets, critical infrastructure, or the spread of disinformation—remains largely unmitigated.

By positioning the Treasury, rather than a dedicated AI regulator, as the de‑facto overseer of AI risk, the administration signals a preference for a leaner regulatory architecture. That choice could shape the broader governance landscape for years, influencing whether future legislation will grant the Treasury explicit authority over AI or leave oversight to existing agencies like the Commerce Department.

Moreover, Bessent’s reference to the record‑low poverty figure ties AI safety to the administration’s broader narrative of economic success. If lawmakers accept that the Treasury can simultaneously manage bond markets, curb inflation, and supervise AI risk, they may be less inclined to endorse a separate, possibly more intrusive, AI agency.

What the sources show

All six outlets report that Bessent defended the voluntary safety stance, but they emphasize different facets. Yahoo Finance and Politico foreground the “stop‑button” claim, presenting it as the centerpiece of his argument. Reuters adds a diplomatic dimension, noting the upcoming meeting with a Chinese official—a detail absent from the other reports.

The Hill and finance.biggo.com focus on the macro‑economic backdrop, highlighting Bessent’s citation of the administration’s poverty record and the Treasury’s bond‑buying program as evidence of effective risk management. CNBC frames the hearing as a broader contest over the Treasury’s interventionist policies, suggesting that Bessent’s AI remarks were part of a larger defense of the Trump economic legacy.

Only Politico provides a verbatim paraphrase of Bessent’s “could stop any time they want to” line, and no source supplies quantitative data on AI‑related incidents, compliance rates, or the efficacy of the voluntary framework. The absence of hard metrics leaves a gap that lawmakers are likely to press in future sessions.

What’s next

The committee has scheduled a follow‑up hearing for early October to review written responses from the Treasury and to hear testimony from senior officials at the Department of Commerce, which is drafting the next iteration of AI policy guidance Reuters. Lawmakers have also requested that industry groups submit additional voluntary best‑practice documents by the end of September, a deadline that will test Bessent’s assertion that the “stop‑button” approach is sufficient.

Mid‑November, Bessent is expected to travel to China for the meeting with the senior official he referenced. Observers will watch for any joint communiqué that could signal the first formal step toward international coordination on AI safety standards.

In parallel, the Treasury will release its quarterly bond‑market outlook next month. Analysts anticipate that the report will address how AI‑driven trading algorithms affect market liquidity, a point Bessent hinted could intersect with the broader safety conversation The Hill.

Should industry submissions fail to address transparency and auditability concerns, the House may push for mandatory reporting requirements, potentially shifting the policy debate toward formal regulation. Conversely, a robust set of voluntary guidelines could reinforce the administration’s case for a lighter‑touch approach.