UpTrajectory Review

Lina Khan, who recently completed her tenure as Federal Trade Commission chair, has escalated her critique of artificial intelligence industry leadership by calling for criminal prosecution of AI company CEOs. Her argument rests on a 1934 precedent involving the prosecution of utility executives, suggesting that the same legal framework could apply to technology executives whose products cause widespread consumer harm. This represents a significant rhetorical shift from the regulatory and civil enforcement approach that characterized her own tenure at the FTC, where she pursued aggressive antitrust cases against Amazon and Meta but stopped short of advocating criminal liability for corporate leadership.

For small-business operators, Khan's proposal carries immediate practical weight regardless of its political feasibility. Most businesses now rely on AI tools for customer service, hiring, pricing, or marketing—systems they purchase from vendors whose liability terms are typically buried in clickwrap agreements. If criminal liability for AI harms becomes even a marginal political possibility, vendor contracts, insurance requirements, and compliance documentation will shift dramatically. A bakery using an AI scheduling tool or a contractor relying on automated lead scoring could find themselves caught between vendor disclaimers and new expectations of due diligence they are ill-equipped to meet.

What warrants skepticism here is the historical analogy itself. The 1934 utility prosecutions occurred in a radically different regulatory environment, with state-granted monopoly franchises and direct rate-setting authority that created clear lines of accountability. AI CEOs operate in a fragmented market with global competition and First Amendment protections for algorithmic outputs that did not constrain utility pricing. Khan's framing also elides a tension in her own record: she had four years to test criminal referrals at the FTC and did not pursue them, suggesting either that the legal theory is weaker than presented or that interagency coordination barriers are insurmountable. The Register's coverage does not interrogate this gap.

The downstream effects split unevenly across the business landscape. Large enterprises with dedicated compliance and government relations functions would likely capture the regulatory process, shaping criminal liability standards to their advantage while smaller competitors absorb disproportionate compliance costs. Meanwhile, AI vendors would accelerate geographic arbitrage—relocating operations, incorporation, or data processing to jurisdictions without extraterritorial criminal exposure. The paradox is that Khan's populist framing could produce concentrated market power, the very outcome her antitrust work sought to prevent. Insurance markets for AI liability remain thin; criminal exposure would freeze them entirely until actuarial models catch up, leaving early adopters self-insuring against risks they cannot quantify.

What to watch is not Khan's proposal itself, which faces near-certain opposition from a Republican-majority FTC and a business-friendly judiciary, but its migration into state-level legislation and private litigation strategies. California's legislature has already shown appetite for AI-specific liability frameworks, and plaintiff attorneys are actively testing novel theories of corporate officer liability. Small-business operators should audit their AI vendor contracts now for indemnification clauses and jurisdiction provisions, document their reliance on vendor representations about safety testing, and monitor whether their industry associations are participating in liability standard-setting or being shut out by larger players. The 1934 precedent may be historically shaky, but the political energy behind it is not going to dissipate.

Takeaway: Audit your AI vendor contracts for indemnification and jurisdiction clauses before liability frameworks harden around you.