Last week I watched a company executive refuse to answer a simple question while a security team scrambled in the wings. You could feel the room tilt toward panic, the reporters sharpening their pens. That’s the kind of moment Lina Khan wants you to see differently.
Last September the FTC opened probes into Meta, xAI and other firms over AI “companions.”
I was on the phone with an engineer who described the agency’s letters as the kind of cold, sober reset the industry rarely gives itself. The inquiry wasn’t theater: it asked how safety-testing was performed, how underage users were protected, and whether emotional-design choices carried harm. You’ve seen the headlines about bots that overstep—this is the regulator following up.
Do existing laws apply to AI?
Yes—and that’s the point Khan is making in plain language. The FTC already enforces against “unfair methods of competition” and defective products; those statutes aren’t labeled “AI-except.” When your model behaves like a product that can hurt people, regulators have tools to act.
A 1934 Supreme Court decision explicitly outlawed a “race to the bottom” that looks a lot like today’s AI panic.
FTC v. R. F. Keppel & Bro., Inc. said companies can’t force rivals to descend into harmful practices just to stay in business. Read that sentence again: a court in 1934 anticipated an industrial dynamic that now crops up in machine-learning labs. When firms pursue dangerous shortcuts because they’re afraid of being outcompeted, antitrust law can be invoked to break that spiral.
CEOs and companies have already been probed together—Amazon, Apple, Google, Meta have all been in the crosshairs.
I’ve spent years watching how investigations shift incentives. When the FTC partners with the Justice Department, the message is blunt: accountability can land at the top. That’s why Khan emphasizes prosecution of the corporate decision-makers who greenlight risky deployments, not some futuristic statute that sits on a shelf.
Can CEOs be prosecuted for AI harms?
The short answer is yes: prosecutors can and have pursued executives when corporate choices foreseeably caused consumer harm. Khan’s point isn’t ambition; it’s pragmatism. If a company knowingly ships a dangerous, unvetted AI product, existing product-liability and competition doctrines are the leash that can be tightened today.
The industry talks about “frontier” models and recursive self-improvement while security incidents like the Hugging Face breach rattle confidence.
That hack and others aren’t just headlines—they’re reminders that technical capability and organizational governance don’t always move together. When advanced models are paired with weak controls, bad outcomes are more likely. You don’t need prophetic legislation to respond to a concrete failure.
Alvaro Bedoya warned that invoking doomsday scenarios can serve as cover for an oligopoly: he called the “AI will wipe out humanity” framing a hyperbolic strawman being used to justify a cartel of rich companies. That’s an authority cue worth heeding. If fear becomes the argument for privatized rulemaking, regulators lose leverage and the public loses oversight.
Could AI companies form a cartel?
They could try. The law already treats coordinated behavior that restrains competition as illegal. Khan’s invocation of historical precedent points at this: antitrust statutes were designed to stop dominant players from freezing the market in place—whether by price-fixing or by erecting technical barriers to competition.
You’ve seen the headlines, the fears, and the think pieces; now consider what enforcement actually looks like on the ground.
Prosecutions and inquiries change incentives faster than new legislative drafts. When enforcement hits a balance sheet, boards pay attention. When executives face real legal exposure, product roadmaps are redrawn. Enforcement is not glamorous, but it’s practical: it forces companies to make safer choices now.
I’ll be blunt: this is not charity. If a CEO signs off on releasing a system that the company knows is dangerous, that is a legal problem comparable to selling a defective physical product. The FTC’s tools are not theoretical—they’ve been used, refined, and paired with Justice Department power. Think of regulators not as traffic cones but as the emergency brake you hope someone pulls before the car barrels over a cliff.
Two clear metaphors help here: the industry is running, in places, like a racecar without brakes, and regulators can act as a lighthouse in a storm—visible, persistent, and capable of changing a captain’s course. These images aren’t pretty; they’re practical.
So when you hear calls for a brand-new legal regime, ask who benefits from delay. Khan’s message is surgical: use what you already have. Use it now. Do you want new laws that sit in committees for years, or real accountability aimed at the people making the decisions today?
Which would you prefer to see: a courtroom summoning a CEO to answer for a dangerous release, or another blueprint promising safety at some distant future date?