In a provocative new proposal, two New York University business professors are calling on Congress to impose strict personal liability on executives and board members of companies developing frontier AI models, arguing that current legal frameworks are inadequate to prevent catastrophic harms from autonomous systems.
Hans Taparia and Bruce Buchanan, writing for The World Signal, invoke a 19th-century legal principle articulated by Oliver Wendell Holmes Jr.: that owners of dangerous animals are liable for any harm those animals cause, regardless of precautions taken. The same logic, they argue, should apply to AI companies that are effectively unleashing unpredictable digital predators on society.
"The possibility of a great danger has the same effect as the probability of a less one, and the law throws the risk of the venture on the person who introduces the peril into the community," Holmes wrote, a quote the professors use to anchor their argument.
The proposal comes amid growing concerns about rogue AI agents, which could potentially disrupt critical infrastructure or even trigger international conflicts. A recent Senate panel has already begun probing the national security threats posed by autonomous AI systems, underscoring the urgency of the issue.
Taparia and Buchanan contrast their approach with that of Nvidia CEO Jensen Huang, who has argued that existing product liability laws are sufficient. "If your product is not ready to ship, don't ship the product," Huang said in a recent interview. But the professors counter that product liability lawsuits target corporations, not individuals, and that directors and officers are typically shielded by indemnification and insurance. Investors, too, enjoy limited liability, capping their exposure at their investment amounts.
They argue that the current incentive structure rewards recklessness: companies racing to beat competitors have little reason to hold back potentially dangerous products, especially when the true risks are poorly understood even by their own developers.
While some, including Bill Gates, have advocated for a federal licensing and oversight agency, Taparia and Buchanan contend that such a body would face an information asymmetry—firms will always know more about their products' risks than regulators, and even the firms themselves lack complete knowledge.
The professors point to the Sarbanes-Oxley Act of 2002 as a precedent. Signed by President George W. Bush after a wave of corporate accounting scandals, that law made CEOs and CFOs personally liable for certifying false financials, with penalties including fines up to $5 million and prison terms up to 20 years. The law led to hundreds of earnings restatements and a marked improvement in financial reporting accuracy.
Taparia and Buchanan's proposal would extend similar personal accountability to the leaders of AI companies. "Imagine if Sam and Dario and Elon and Mark, along with their boards of directors, bore individual liability for the damage their companies may inflict on society," they write. "Imagine if they faced possible prison time for the consequences."
They conclude that such a legal shift would force a more cautious approach to the AI race, keeping "rogue AI agents on a leash." The proposal is likely to spark fierce debate in Washington and Silicon Valley, as policymakers grapple with how to regulate a technology that is advancing faster than the laws designed to govern it.
