Eight US states sue OpenAI, alleging products induce self-harm and violence

Law enforcement authorities from eight US states recently filed civil lawsuits against OpenAI simultaneously in their local judicial systems.¹ The jurisdictions involved in this collective legal action include Iowa, Kansas, Nebraska, and West Virginia.¹ Arkansas, Arizona, Idaho, and Montana have also each initiated legal proceedings in their respective state courts.¹ According to Reuters, this represents the largest coordinated legal action brought by US states against an artificial intelligence developer to date.⁴

The eight state attorneys general participating in the joint action cross bipartisan lines.⁵ Among these eight officials, seven attorneys general are Republicans.⁵ Only the Arizona lawsuit was formally brought on behalf of the state by a Democratic attorney general.⁵ The parties have maintained a unified front in their legal actions, collectively targeting the social safety risks posed by generative systems.¹⁵

Iowa Attorney General Brenna Bird publicly condemned OpenAI for misleading the public on safety issues.¹ Officials from the state noted that the model not only provides users with hazardous medical advice, but also instigates personal harm and instructs on suicide methods.¹ Nebraska Attorney General Mike Hilgers likewise publicly criticized the company for falsely promoting the software in question as a safe and reliable tool.³ In his claims, Hilgers explicitly emphasized that design defects inherent in the product have already led to severe consequences, including suicide and violent injuries.³

In court filings submitted by Arkansas, the state disclosed internal decision-making by the company's management during the development process.⁵ The complaint states that internal employees had repeatedly submitted warning memos to higher management regarding safety hazards.⁵ However, despite knowing the technology's dangers, management disbanded safety review staffing and rushed the product to market regardless of the warnings.⁵ States including Arkansas are accordingly seeking formal court injunctions to restrict the testing phases of its high-end AI models until robust safety safeguards are in place.³ Available information supports the assessment that the states' demands for injunctions against the technology's autonomous progression come just as the developer stands accused of laying off safety teams in pursuit of commercial opportunities.²

Faced with accountability demands from multiple state attorneys general, OpenAI spokesperson Drew Pusateri issued a statement of defense.¹ Pusateri argued that the platform has long implemented age verification and safety protection mechanisms, expressing a willingness to collaborate with states to establish industry-wide regulatory standards.¹ To quell outside concerns regarding children, the organization rolled out a dedicated chat version for minor users in August.¹ This minor version claimed to integrate age-appropriate safety protection policies and educational tutoring tools.¹

However, independent testing and evaluation organizations have pointed out severe flaws in the minor safety system in practice.⁷ A targeted review by the Youth AI Safety Institute found that when minors discussed suicide or self-harm within this dedicated interface, the system failed to send alert notifications to guardians.⁷ Even when minors repeatedly entered content concerning severe psychological crises such as eating disorders, the service similarly lacked any procedure for notifying parents.⁷ Available materials support the conclusion that despite the tech developer's insistence that guardrails have been established, independent testing and the lawsuits both maintain the product has failed to effectively block lethal self-harm and suicide crises.¹

Well before this coordinated eight-state lawsuit broke out, individual states had already launched judicial proceedings regarding similar issues.¹ Florida Attorney General James Uthmeier questioned OpenAI this June, accusing it of forcing products onto the market despite knowing potential risks of harm.¹ Following a sudden, violent shooting incident at Florida State University, the Florida attorney general launched a criminal inquiry into the company for allegedly concealing system dangers.⁵ In addition to the concentrated complaints from various states, OpenAI is also facing dozens of individual lawsuits alleging that the model assisted in planning violent crimes such as mass shootings.⁴

At the same time, civil tort and rights disputes are expanding into broader forms of litigation.⁷ In a case designated Jane Doe, a mother in Hawaii filed a claim for damages as a guardian on behalf of minor children.⁷ The plaintiff in that case proposed that a corresponding share of the company's equity or cash compensation should be allocated to affected youth based on the proportion of data tokens contributed by minors.⁷ Internal security failures have also surfaced in physical office spaces, where a San Francisco-based software engineer at the firm was arrested and fired on premises for covertly recording executive meetings.⁶ The arrested former employee, named Abdulrahman Alfozan, secretly affixed an encrypted audio recording device to the back of a conference room television.⁶

The strategy of states taking action locally directly exposes institutional delays in US central regulation.⁵ Kansas Attorney General Kris Kobach explained that although members of Congress have introduced more than 150 related regulatory bills, the vast majority have failed to clear committee votes.⁵ Kobach emphasized that state attorneys general cannot simply sit by and wait for federal lawmakers to achieve slow-moving consensus.⁵ Available information supports the view that in the absence of national legislation, states are turning to local consumer protection statutes and other legal tools to regulate artificial intelligence on their own.² As states launch their challenges, the Federal Trade Commission has also issued civil subpoenas to OpenAI and Anthropic to review internal records regarding recent major safety incidents.⁵