ABSTRACT
Powerful AI systems stand to improve our lives dramatically. They also pose substantial risks. Many experts warn that frontier AI systems could cause large-scale, catastrophic events, either by accident or because of malicious human use. Today, frontier AI firms underinvest in preventing such catastrophes. Regulation could help, but there are three major challenges. First, no single firm is financially capable of covering the full damages of a sufficiently catastrophic event. This judgment proofness means that AI companies do not fully internalize the catastrophic risks they generate. Second, the significant first-mover advantages in the market for cutting edge AI models produces a perverse race dynamic. Investing in safety risks losing ground to competitors who don’t do the same, creating a prisoner’s dilemma in which each firm is incentivized to cut corners, even if all would prefer to be safer. And third, traditional regulatory approaches struggle to address these issues because the speed and volatility of AI development outstrip the state’s ability to monitor, understand, and regulate it effectively.
This Article proposes a solution to these challenges: shared residual liability for frontier AI companies. Modeled after state guaranty insurance funds, the proposed regime would hold the most powerful AI companies jointly liable for catastrophic damages in excess of individual firms’ ability to pay. This approach would mitigate the judgment proof problem by increasing the funds available for victims and, therefore, the amount of risk the industry collectively internalizes. It plausibly would mitigate the perverse race dynamic by tying each firm’s financial fate to the behavior of its peers, incentivizing firms to monitor, pressure, and collaborate with one another to promote safer practices. Finally, relying on liability to incentivize firms to identify and undertake effective precautions, shared residual liability might also mitigate regulation’s pacing problem, leveraging AI firms’ superior information about how optimally to avoid AI catastrophe.
Friedman, Benjamin Gil, Shared Residual Liability for Frontier AI Firms (July 5, 2025).
Leave a Reply