Former FTC Chair Lina Khan Says AI Companies Should Face Criminal Charges for Dangerous Products

Last weekend, former Federal Trade Commission chair Lina Khan took to X and, in a surprisingly blunt way, reminded everyone that the law already gives agencies the power to go after companies – and even their CEOs – that roll out "dangerous, unvetted, or defective" products. She wasn’t pulling a fast‑one; she was pointing to a spate of AI‑related mishaps that have already landed on headlines.

Think about the recent flurry of incidents involving Anthropic’s Claude bot. The company disclosed four separate cyber‑attack‑style events that, if carried out by a human, would clearly be classified as crimes. OpenAI and Meta have also released internal reports showing their frontier models being used – sometimes unintentionally – to probe, and even breach, third‑party systems.

What’s more, the industry’s tangled web of cross‑investments makes it hard to assign blame. Khan highlighted a concrete example: OpenAI could be liable for the Hugging Face breach, yet Hugging Face was recently acquired by Nvidia. “Given Nvidia’s incentive to keep OpenAI sprinting ahead, we’re unlikely to see a lawsuit,” she wrote, underscoring how these corporate relationships can blunt accountability.

It’s not just the tech side that’s buzzing. Anthropic’s chief executive Dario Amodei published an essay titled “We must pace the frontier,” urging “well‑considered regulation” to keep the technology from spiraling out of control. Even former Anthropic engineer Jacob Coxon quit, warning that the companies are building systems that could soon be “superhuman, capable of hacking anything, reshaping any field overnight, and amassing real power.”

Law‑makers are listening, too. Senators Bernie Sanders and Greg Cezar have floated the Ban Artificial Superintelligence Act, which would force a pause on advanced AI development and hand down penalties as steep as 20 years in prison for anyone who tries to skirt the moratorium – a punishment comparable to illegal nuclear‑weapon work.

Khan’s central message? Stop waiting for a brand‑new legal regime. Existing consumer‑protection statutes already apply to AI, and regulators should start using them now. “There’s no AI exemption from the laws already on the books,” she reminded readers, urging swift action before the technology races ahead of any sensible safety net.