US President Donald Trump said on September 21 that the Justice Department and other law-enforcement agencies could impose restrictions on the artificial intelligence industry if necessary, while again backing continued AI development. The comments put the boundaries of the administration's approach to frontier-AI regulation back in focus, including the potential legal liability of developers.
Trump Leaves Door Open to Justice Department Action
In a post on Truth Social, Trump said the US government would remain cautious and that the Justice Department and other law-enforcement agencies could constrain the sector "if necessary." He did not specify what measures might be taken or identify any AI company as the subject of a planned enforcement action.
At the same time, Trump repeated his longstanding preference for a lighter-touch approach to AI. He said he would only "encourage artificial intelligence" and specifically referred to "super intelligence," or SI. The comments did not amount to an announcement of a broad regulatory tightening. Instead, they kept the focus on when the government might intervene and how much responsibility companies should bear for risks linked to their systems.
Bessent Says Frontier Labs Will Not Get Liability Protection
Treasury Secretary Scott Bessent said the same day that the Trump administration would not provide frontier-AI laboratories with protection from liability for related risks. Bessent said the laboratories would have to bear responsibility themselves and could slow their development programs at any time if they believed progress was moving too quickly.
Liability is one of the central issues in the US AI policy debate. As models become more capable, companies may face legal and commercial exposure related to product safety, inaccurate outputs, data use and potential wider social effects. Bessent's position indicates that, at least on liability, the administration has not signaled plans for a broad federal backstop for AI companies.
Anthropic CEO Calls for Federal Rules
Anthropic Chief Executive Dario Amodei has renewed his call for the US government to establish a federal framework for AI regulation. He has argued that frontier-AI laboratories should moderate the pace of development and use the time gained to improve the safety of their most advanced models.
Some of Amodei's views have been supported by OpenAI Chief Executive Sam Altman and SpaceX Chief Executive Elon Musk. Their shared concern is whether frontier laboratories should establish clearer safety assessments and risk controls before continuing to expand model capabilities. Trump and Vice President JD Vance, however, have previously expressed skepticism toward AI companies' calls for government rules that could slow the industry.
States Move Ahead as Federal Authority Remains Contested
With Congress still unable to agree on a unified set of AI rules, states including California and Illinois have advanced their own regulatory initiatives. The Trump administration has also issued several AI-related executive actions, but presidential authority cannot substitute for legislation from Congress. The scope and enforcement of any federal rules therefore remain subject to limits.
In his latest post, Trump called the current debate over AI safety a "scam" and linked it with issues including climate change, Russia and Ukraine. On September 17, he said the only AI "guardrail" the United States needed was a president who was sufficiently strong and intelligent, adding that he could manage the technology.
Safety Dispute Continues to Shape Policy Debate
Discussion of the possibility that AI could pose extreme risks to humanity intensified after former Anthropic researcher Jacob Kosen resigned. In his resignation statement, Kosen warned that some employees at AI laboratories privately feared the technology could eventually create risks to human survival.
The administration is therefore presenting two positions at once: it says judicial and law-enforcement tools could be used to constrain AI when necessary, while continuing to emphasize technological development. The Treasury Department has also made clear that frontier laboratories should not expect immunity from liability. For developers and investors, the unresolved questions are whether federal oversight will become more specific and how much responsibility companies will bear for model safety and potential harm.