Former FTC chair Lina Khan: hold AI CEOs personally liable, the way 1934 did to Wall Street

Former FTC chair Lina Khan says U.S. prosecutors do not need a new AI law to put the people running AI companies in personal legal jeopardy. The tools to do it, she argues, were written in the wreckage of the 1929 stock market crash, and they have been waiting on the shelf ever since.

Khan laid out the argument in a lengthy X post on Sunday and at the AI Seoul Summit earlier this month, after a string of well-publicized incidents in which AI products from Anthropic, OpenAI, and others caused criminal or near-criminal harm. Her core claim: there is no AI exemption from laws already on the books, including a 1934 statute built to make individual financial executives personally criminally liable for what they sold to the public.

The mechanism is the Securities Exchange Act of 1934, the post-crash reform that created the Securities and Exchange Commission and wrote personal criminal liability into U.S. securities law. The most-cited hook is Section 10(b) of the act and the SEC's Rule 10b-5, which together make it a felony for an individual executive to make materially false or misleading statements in connection with the sale of a security. The original target was Wall Street salesmen who sold products they knew could not deliver. Khan's pitch is that an AI company CEO who tells the public their model is safe, aligned, or production-ready falls inside the same legal reach when those claims are material to investors and end users.

She is not stopping at the 1934 securities hook. Her X post builds a five-part case. The first three parts stay inside consumer-protection and competition law, the kind the FTC enforced during her 2021-2025 tenure: shipping an unvetted AI model as an "unfair or deceptive" practice under Section 5 of the FTC Act, using market power to extract competitively sensitive data from customers as an "unfair method of competition," and using opaque cross-investments across the AI stack to evade scrutiny. The fourth invokes data-security law, including the FTC's 2022 settlement with Drizly that held CEO James Cory Rellas personally responsible for the company's breach.

What Khan is specifically asking for is a shift in who pays. Today, when an AI product causes harm, the bill tends to land on the corporation through consent decrees, fines, or shareholder litigation. Khan wants the person in the corner office named in the charging document. That is the 1934 template, applied to a 21st-century product.

The 2026 enforcement record is already moving in that direction. In April, the U.S. Attorney's Office for the Southern District of New York announced a guilty plea from the founder and former CEO of an AI company on securities fraud charges, applying the 1934 hook to an individual AI executive. Weeks later, Reuters reported that the former CEO and former CFO of a now-bankrupt AI company had been charged with fraud in the same posture. Both cases sit on the Section 10(b) rail Khan is pointing to.

The bottleneck, on her reading, is not the statute. It is the will to use it. The current AI accountability system, she writes, mostly runs at the company level, with individual executives insulated by the corporate veil and by the network of cross-investments that ties the major labs to their would-be regulators. The Independent noted Khan's example: OpenAI could face liability over the Hugging Face incident, but Nvidia's purchase of Hugging Face means the suit is unlikely to be filed, because Nvidia has a strong interest in seeing OpenAI keep shipping.

Whether the next AI-CEO-level charging document comes from the SEC, DOJ, or a state attorney general is the open question. Khan's argument is that the only thing missing is a decision to file.