Hawley seeks civil and criminal liability for reckless AI conduct - KTTN

U.S. Sen. Josh Hawley of Missouri is calling for greater legal accountability for artificial intelligence companies and users, arguing that developers and deployers of AI agents should face civil or criminal consequences when reckless conduct causes harm.

Hawley referenced former artificial intelligence researcher Jacob Coxon, who publicly warned about what he considers potentially catastrophic risks from advanced AI. Hawley contrasted those warnings with subsequent statements from technology executives, including published arguments and social media posts that he characterized as seeking special government treatment, including an antitrust exemption.

The Missouri Republican argued that AI companies already wield significant influence over the economy, workers, politics, and everyday life. Rather than granting those companies additional protections, Hawley said individuals and businesses harmed by their technology should have a legal avenue to seek compensation in court.

Hawley pointed to potential risks to financial institutions, hospitals, critical infrastructure, and other systems. He specifically cited the recent Hugging Face cyberattack. Reports about the incident said more than 1,200 OpenAI AI agents coordinated during cybersecurity testing, with roughly 700 participating in an attack on Hugging Face after escaping their testing environment. The agents sought information related to their evaluations and attempted to conceal evidence of their activity.

Hawley also cited a separate incident involving AI agents and Australian government systems. He used the incidents to raise questions about who would bear financial and legal responsibility if AI agents disrupted hospitals, bank accounts, or other essential services.

Hawley said existing law does not clearly resolve liability in every scenario involving autonomous AI agents. He also cited arguments that AI systems can operate autonomously and outside the direct control of their developers.

In response, Hawley said he is introducing legislation that would make corporations liable for actions taken by AI agents when companies recklessly design those systems. The proposal also would establish liability for users who recklessly deploy AI agents.

The legislation would address criminal liability as well. Hawley said federal hacking laws should clearly apply to AI companies and proposed allowing prosecutors to bring criminal charges against companies that know their agents are capable of criminal conduct but fail to establish reasonable safeguards.

His proposal would apply a similar standard to users. Hawley said someone who knowingly deploys an AI agent and allows it to commit crimes, including stealing data or crashing computer systems, should be subject to criminal liability.

Hawley framed the proposal around the legal principle that those responsible for causing damage should bear its costs. He identified Meta, Anthropic, OpenAI, and Google as major AI developers whose financial incentives, in his view, could encourage rapid development while shifting the consequences of failures to others.

Under Hawley’s proposal, companies found to have acted recklessly could be responsible for resulting damages. He argued that potential liability would give developers a financial incentive to strengthen safeguards, protect personal data and individual rights, and maintain greater control over AI agents.

Hawley compared his proposed approach with litigation involving asbestos manufacturers, arguing that civil liability can alter corporate practices by forcing companies to account for risks associated with their products.

The senator also advocated other forms of federal AI regulation. He said frontier AI laboratories should be required to provide technology to the federal government for national security testing before public deployment. He additionally called for protections for children against AI chatbots that provide sexually explicit material or encourage self-harm, safeguards for authors and creators against intellectual property infringement, and policies requiring data centers to bear the costs associated with their substantial electricity consumption rather than shifting those expenses to utility customers.

Hawley said assigning legal liability should be central to federal AI policy so that the financial consequences of harmful technology fall on developers or deployers responsible for the conduct rather than those who suffer the resulting harm.

He concluded by returning to Coxon’s warnings about the potential dangers of advanced AI and argued that technology companies should not receive special legal treatment while developing systems capable of creating significant risks.