Khan Calls for Criminal Charges Against AI CEOs, Citing 1934 Securities Law

Lina Khan wants AI executives to face prison time, not just fines. The former chair of the Federal Trade Commission argued that prosecutors should borrow the criminal-liability framework Congress wrote into the Securities Exchange Act of 1934 and apply it to the people who sign off on claims about AI models, according to remarks reported by The Register. Under her proposal, executives who misrepresent what their systems can do — or how safely they perform — would carry personal legal risk rather than leaving the company to absorb a civil penalty.

The 1934 analogy is structural, not rhetorical. That statute lets the government charge senior officers directly when a company misstates material facts to investors, a power that put executives behind bars after the savings-and-loan crisis and again after Enron. Khan's argument is that AI labs make similar representations about capabilities, safety testing and system limits, and that the executives signing those representations should face comparable exposure. Today, most US AI enforcement runs through the FTC's civil authority, producing consent orders and monetary settlements that rarely touch anyone's liberty.

That gap defined Khan's tenure at the FTC from 2021 to 2025. She brought civil actions against Meta, Amazon and Microsoft and rewrote the merger guidelines, but told Congress repeatedly that civil tools move too slowly and cost too little to discipline firms valued in the trillions. Criminal exposure for named officers would change the internal calculus at AI labs, where safety commitments are frequently drafted by the same executives who stand to gain from relaxing them.

The proposal lands as federal AI oversight remains unresolved. The Trump administration has favored deregulation and voluntary commitments, while states including Colorado and California have advanced their own rules. Khan's framing offers aggressive state attorneys general a federal template to borrow, and gives plaintiffs' lawyers a theory to test in civil litigation even without new legislation from Congress.

Industry reaction has split along familiar lines. Trade groups contend that criminal liability would drive engineers out of frontier labs and push talent to competitors overseas, while safety researchers counter that voluntary pledges have already buckled under commercial pressure. Neither side has produced evidence on how criminal exposure would affect hiring or model release schedules, leaving the argument resting on assertion.

What to watch is whether any prosecutor actually tests the theory. No AI executive has been charged criminally over model harms, and no pending case appears to be building toward one. The nearer-term question is whether Khan's argument reshapes the state-level bills now moving through legislatures — or remains a provocation that never leaves the op-ed page.