Trump Says DOJ Will Rein In AI 'If We Have To' After 30-Day Reviews

For general counsel and compliance officers, Trump's latest Truth Social post reads as both a promise of restraint and a warning of discretionary enforcement. The September 21 statement offers no statutory anchor for the DOJ's authority, instead invoking 'tremendous CRIMINAL and REGULATORY power' that the White House has not defined. That gap between voluntary frontier-model reviews and potential DOJ action is the central legal risk to manage.

President Donald Trump on Monday signaled that the Department of Justice and other U.S. law enforcement agencies stand ready to act as a backstop against artificial-intelligence risks, even as he reiterated an aggressively pro-innovation stance that places voluntary corporate cooperation ahead of new statutory constraints. In a Truth Social post dated September 21, 2026, Trump wrote that 'we will be careful' and that the DOJ and other law-enforcement bodies 'will rein things in if we have to,' before adding that he will 'only encourage AI or, SI (SUPER INTELLIGENCE)!' The statement, reported by CNBC and Proactive, arrives amid escalating tension between frontier AI labs and their own researchers over the pace of advanced model development.

“

President Donald Trump on Monday signaled that the Department of Justice and other U.S.

The immediate trigger was the resignation of Jacob Coxon, an AI researcher at Anthropic, who publicly warned that Anthropic and OpenAI are 'gambling with our lives.' That warning has been amplified by an unusual coalition of industry leaders—Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, SpaceX CEO Elon Musk, and Google DeepMind chair Demis Hassabis—who have called for a slowdown in AI development and additional federal oversight. Trump, however, has spent the past two weeks downplaying AI dangers, at times calling them a hoax, and his latest post attempts to reconcile safety rhetoric with a full-throttle competitiveness agenda. He emphasized that the U.S. is 'leading now over China, and everyone else' and that AI is 'bigger than the Industrial Revolution, or the Internet, itself.'

That voluntarism is embodied in Trump's June 2026 executive order, which laid out a deliberately loose review plan. The order asked companies to work voluntarily with the administration to evaluate 'covered frontier models' and provide government access up to 30 days before a broad release. There is no indication that any of those provisions are mandatory, and the White House has largely declined to press for binding legislation. The DOJ backstop therefore occupies an ambiguous middle ground: Trump claims the federal government already possesses 'tremendous CRIMINAL and REGULATORY power' over AI companies—a point he made in a September 14 Truth Social post—but he has not specified which existing statutes or regulatory authorities the DOJ would invoke.

For legal observers, the statement raises more questions than it answers. The Department of Justice does not have a dedicated AI-enforcement statute; its levers are general criminal and civil enforcement tools, including fraud, consumer protection, national security, and civil rights statutes, any of which could be applied to AI-related misconduct on a case-by-case basis. Trump's reference to 'CRIMINAL and REGULATORY power' suggests a posture that treats existing law as sufficient, avoiding the need for new legislation but relying on prosecutorial discretion. That posture creates uncertainty for companies trying to build compliance programs, since the precise triggers for DOJ intervention remain undefined, and an enforcement action could be motivated as much by political signaling as by legal precedent.