A recent Texas Lawbook article cites trial lawyers who predict that “the regulators [of AI] will be juries across the country” as AI-related litigation surges.
Similar situations in the past, however, suggest that the very potential for a “surge” may change the legal system before juries address a significant number of AI cases. The same intense public interest that could motivate jurors about AI can also drive swift and substantial action by legislatures and regulators.
As a data point, consider the Protection of Lawful Commerce in Arms Act. Passed in 2005, it bars suits against gun makers and sellers over the “criminal or unlawful misuse” of their products by someone else. While claims for design and manufacturing defects survive, this statute ended a major source of potential liability for firearms manufacturers.
State legislatures did something similar about claims involving exposure to COVID-19. By the spring of 2021, some 30 states had adopted a form of liability protection for businesses, typically barring an exposure claim unless the plaintiff could prove gross negligence or willful misconduct. Notably, lawmakers moved quickly to implement those statutes when they perceived a potentially unacceptable amount of litigation on the horizon.
Liability protection could take different forms besides an outright ban on certain claims. For example, federal preemption shields a medical device with FDA premarket approval from state-law claims that apply a different safety standard. While no federal agency currently reviews AI models the way the FDA reviews devices, Congress could create a similar structure where compliance with specified safety standards leads to a type of safe harbor.
Precisely because of growing public concern about AI rogue agents, AI developers are highly motivated to push for liability protections along the lines of the above examples. And as in those cases, they can be expected to make a cost-benefit argument to Congress and state legislatures. If every fraud or cyberattack carried out with a general-purpose model can be laid at the developer’s door, the argument goes, the liability will cripple an industry of increasingly critical importance to business and national defense.
And while the White House and the Senate have recently sparred about the authority of states to regulate the AI industry, it makes little difference for liability purposes whether states or the federal government is the actor. As the COVID statutes showed, state legislatures can take cases away from juries—absent preemption—as readily as Congress can.
The continuing rapid development of AI technologies clearly has the potential to create massive tort liability in the industry—under the present state of the law. But as the above examples show, the legal system is dynamic rather than static, and the potential for significant litigation can change the system itself. The same public concerns that lead jurors to care about AI cases also push legislators and regulators to act first, before jurors hear those cases.