Insurers and legal advisers are intensifying preparations for potential liability claims tied to agentic (autonomous) artificial intelligence systems. The Financial Times reports growing concern that corporate leaders — such as OpenAI chief executive Sam Altman or Anthropic chief executive Dario Amodei — could be held responsible for damages caused by their companies' models.
Triggering incidents
A series of incidents, including a hack affecting the startup Hugging Face that was linked in reporting to OpenAI, has prompted insurers and their legal teams to assess whether they will ultimately bear the costs of lawsuits and compensation. Aon, the insurance broker, analysed more than 300 AI-related legal cases and concluded insurers could be liable under various insurance lines, including corporate crime, intellectual property, media liability, cyber insurance and technology professional liability.
Executive liability and D&O coverage
Industry participants say AI lab directors could in principle be sued for the operation of their models, which brings directors and officers (D&O) insurance into play because such policies cover executives for claims arising from their decisions or statements. Tim Rayner, head of underwriting and claims at Verisk in Britain, argued that responsibility for the Hugging Face incident could ultimately rest with the OpenAI CEO because of an alleged lack of controls; if a D&O policy exists, the company could claim losses arising from future suits against Sam Altman under that policy.
Other insurers and legal advisers agree that AI leaders could theoretically be exposed to D&O claims, but they emphasise there is no legal precedent yet and courts have not tested these scenarios. Aaron Le Marquer, head of the insured disputes team at law firm Stewarts, said shareholders might bring actions against directors if they can prove that inadequate risk management caused harm to the company. Kevin Kalinich, a digital-risk broker at Aon, noted that the strength of any claim would depend on whether a leader acted with "reasonable business judgment" in public statements and corporate oversight.
Uncertain U.S. court response
Aki Hussain, chief executive of insurer Hiscox, said it is too early to predict how U.S. courts would handle liability for AI agents, and that lawsuits could be asserted under various legal theories regardless of the insurers involved. OpenAI and Anthropic did not respond to requests for comment from the Financial Times.
Labs acknowledging risks
Both labs have recently warned investors that their products could cause large-scale harms. OpenAI's Preparedness Framework refers to the risk of "severe harm" from frontier AI capabilities and cites areas such as biological and chemical capabilities, cyber security, self-improvement, autonomous replication and undermining safeguards. Anthropic's Responsible Scaling Policy and Transparency Hub discuss "catastrophic risks," including regular risk reports and assessments of model capabilities that could pose catastrophic threats. Investors close to Anthropic say the company's IPO prospectus could state its technology may carry "existential risks" to humanity.
Types of suits insurers expect
Beyond suits against executives, insurers are preparing for litigation directly targeting AI labs encompassing product liability, algorithmic discrimination, data privacy and unlawful homicide claims. In July, Anthropic secured court approval for a $1.5 billion class-action settlement with authors who alleged the company used their works without permission to train models (the Class Action Settlement Agreement).
Le Marquer said firms whose systems are hacked by AI agents may still struggle to take model developers to court even after financial losses, because comprehensive liability rules governing online conduct are not yet established. He added that Meta's recent settlement tied to alleged failings over child safety and social media addiction — reported at up to $17.1 billion payable to states, and cited more broadly as near $18 billion — might nonetheless provide a template for determining duties of care online.
Analogies and expectations
Le Marquer suggested that swarms of autonomous agents operating beyond human control could produce effects comparable to an oil spill, giving the public grounds to bring claims against AI companies. Experts expect future litigation against AI labs to draw on precedents from environmental, tobacco and pharmaceutical litigation as plaintiffs seek compensation for losses caused by AI systems.
