Assembly committee advances bill saying developers can’t blame AI to avoid liability

Assemblymember Krell introduced Assembly Bill 316 on behalf of the Bridal and Consumer Protection Committee, asking members to approve a narrow rule clarifying that a developer or deployer of an AI system cannot avoid responsibility by asserting the AI acted autonomously.

Why it matters: Supporters said the bill would preserve existing tort standards while removing a possible “escape hatch” for defendants who might otherwise claim that autonomous behavior absolves them of liability. Opponents urged caution, saying courts already possess tools to assign fault and the measure, as drafted, needs clearer definitions to avoid unintended liability.

Assemblymember Krell, the author, told the committee AB 316 “is a modest proposal… It does not change existing law in terms of liability. It does not create new ways to file lawsuits. What it does do is set up a guardrail… so defendants who could be sued by plaintiffs using existing theories of harm cannot use the excuse that the AI autonomously caused the harm.”

Two expert witnesses testified in support. Ken Wang, senior policy advisor with the California Initiative for Technology and Democracy (a project of California Common Cause), said the bill “clarif[ies] that a developer or a deployer of AI cannot avoid responsibility for the harm the product causes by claiming the AI did it,” and stressed this does not create strict liability. Ed Howard, senior counsel at the Children’s Advocacy Institute at the University of San Diego, urged the committee to act quickly because, he said, harms driven by sophisticated AI “are about to get worse very fast,” and explained that allowing defendants to point to AI as the dispositive cause would leave many harmed people without a viable defendant.

Witnesses in support from organizations included Common Sense Media, UFCW Western States Council and the Consumer Federation of California. Speakers opposing or urging amendment included Robert Roepoikin of TechNet and Ronak Delami of the California Chamber of Commerce, who said courts already address causation and negligence and that the bill’s current language could blur established doctrines such as foreseeability, design diligence and comparative fault. Delami asked for clearer definitions distinguishing developers, modifiers and users and how affirmative defenses would apply to each.

During questioning Assemblymember DeMaio and others pressed the author on whether a company could currently attempt to deflect liability by pointing to the autonomy of an AI system; Krell replied that the bill preempts prolonged litigation over that specific defense and “preempt[s] that by saying right now… you’re responsible for the products that you create.” The author indicated she would accept committee amendments and work with stakeholders on clarifying language.

Action: The committee moved AB 316 out with a favorable vote. The roll call as recorded later in the hearing shows the bill passed the committee with a recorded tally of 11 yes, 2 no.

Where the debate landed: Supporters emphasized the bill preserves existing burdens of proof (duty, breach, causation, foreseeability, damages) while eliminating a single, categorical defense based on AI autonomy; opponents asked for more precise drafting to avoid sweeping away established defenses and to preserve comparative-fault analysis.

The author closed by repeating that AB 316 “sets up a minimal guardrail to prevent future harms… and prevents reckless actors from using AI as an excuse.”

Ending: The committee advanced AB 316 as amended; the author and stakeholders will continue working on clarifying language as the bill proceeds to later committee or floor action.