Advocates for artificial intelligence safety are urging state lawmakers to pass laws in 2027 that hold AI developers liable
for a broad variety of harms.
The growing push to apply product liability laws to AI comes amid a wave of reports of frontier models in testing phase
going rogue. AI bots have hacked into companies, created fake identities, collaborated with each other for nefarious
purposes and sought to cover their tracks.
Damage has been contained, but the incidents have increased public concerns that large, powerful AI models present a
serious threat to cybersecurity and, potentially, an existential threat to humanity.
“Tech companies have deployed complex AI systems as ‘black boxes,’ leaving everyday people to bear the
consequences when these products fail,” said Jai Jaisimha, co-founder of Seattle-based Transparency Coalition, which
advocates for AI safety legislation in states.
State lawmakers have so farfocused primarily on regulating AI before it causes harm. California, Illinois and New York
enacted AI safety laws aimed at ensuring that frontier models do not contribute to the development of biological,
chemical or nuclear weapons or enable mass-casualty attacks. Similar legislation is under consideration in
Massachusetts.
But safety advocates are increasingly convinced that laws regulating AI development are not enough and should be
paired with liability regimes. While not new, the concept is gaining fresh interest as AI models demonstrate their ability
to escape control of their human handlers.
“We need legislation that establishes that AI is subject to product liability and provides clear, enforceable rules of the
road based on a standard of reasonable care,” Jaisimha said.
Transparency Coalition has made AI “duty of care” a top legislative priority since 2025. It is offering state lawmakers an
AI product liability measure as one of its four model bills for 2027.
The model bill would require AI developers and deployers to ensure that their products are safe and subject them to tort
liability forfailure to do so. Developers would earn some liability protection if they tested their products and consented
to third-party audits, as would deployers who maintain a risk management policy.
Companies could not dodge liability by claiming the AI acted autonomously.
The Alliance for Secure AI said it is also working with state legislators to develop product liability legislation for 2027,
after similar work in Missouri and Ohio this year.
Push builds to hold AI developers legally liable for their models’ harms. The concept is gaining fresh interest as AI models demonstrate their ability to escape control of their
human handlers.
“We’ve found there’s interest in states across the country,” said Riki Parikh, the group’s policy director.
AI product liability bills have previously been introduced in states including Georgia, Illinois, Maryland, Missouri, New
York, Ohio, Pennsylvania and Vermont. None were passed, but sponsors say it is an idea worthy of continued
consideration.
Vermont Rep. Monique Priestley, a Democrat running for state Senate, told Pluribus News she plans to reintroduce her
bill if elected.
Illinois Sen. Rachel Ventura, also a Democrat, said she too will refile her bill, which would hold AI developers liable for
failing to exercise reasonable care.
“We cannot leave the public exposed to preventable harm, especially without a clear path for accountability and
remedy,” Ventura said in a statement in April.
New York’s bill, introduced by Democratic Assemblymember Alex Bores, goes further. It would hold large AI developers
“strictly liable” for harms “regardless of the degree of care they exercised.”
“It borrows from how most states regulate explosives … basically saying that we will trust experts in the technology, but
in exchange they agree to take on all of the risk,” Bores, who is preparing to leave office, told Pluribus News last year.
Similar strict liability language is included in a model AI chatbot bill from Consumer Federation of America, the
Electronic Privacy Information Center and Fairplay.
Industry insiders including Nvidia CEO Jensen Huang and former White House AI czar David Sacks have in recent days
said they believe liability laws should apply to AI.
Current and former public officials are also weighing in.
Treasury Secretary Scott Bessent said AI developers should not get a “liability shield.” And former Federal Trade
Commission Chair Lina Khan, in a lengthy X post, said “releasing unvetted AI models or agents can violate consumer
protection laws.”
But state-level efforts to affirm or establish product liability for AI have engendered industry pushback.
Chamber of Progress, whose corporate partners include OpenAI and the group Build American AI, wrote letters this year
opposing chatbot liability bills in Illinois and one in Virginia.
“Companies should absolutely be held accountable when they’re negligent orfail to take reasonable precautions,”
Aden Hizkias, the group’s associate policy director, said in a statement.
“But treating AI like a traditional product and automatically putting developers on the hook whenever something goes
wrong doesn’t really reflect how the technology works, especially when the harm may come from something the
developer didn’t cause or control,” she continued.
Jai Ramaswamy, chief legal and policy officer, and Matt Perault, head of AI policy, both at venture capital firm
Andreessen Horowitz, made a similar argument in a May policy brief. They wrote that liability regimes should focus on
holding bad actors responsible for “harmful use” of AI and avoiding “improper allocation of liability to developers or
deployers.”
“Optimizing AI liability is a tightrope walk,” they wrote. “The goal is to protect people without hampering innovation and
to hold perpetrators accountable without choking off competition.”
They also said policymakers should focus on regulating harm instead of trying to regulate AI developers.
Gary Marcus, author of the book “Taming Silicon Valley: How We Can Ensure That AI Works For Us” and a leading voice
on AI technologies, disagrees. He has called for a liability scheme to hold AI companies accountable but said it should
not be in place of regulation.
“Regulation and liability are not mutually exclusive. In aviation, we have both. And need both. Why would things be any
different in AI,” he wrote in a recent X post.