OpenAI and Anthropic are weighing the use of investor capital to settle or safeguard against potential multibillion-dollar legal claims, as insurance companies shy away from offering full coverage for the growing risks tied to artificial intelligence, according to the Financial Times. The two U.S.-based AI startups are facing mounting legal challenges while navigating an insurance market that has yet to adapt to the scale and complexity of emerging AI liabilities.
Both companies maintain standard business insurance policies, but industry experts told the Financial Times that AI model providers are likely to struggle to obtain protection that fully reflects their exposure. OpenAI, which has enlisted Aon, the world’s second-largest insurance broker, to help structure its coverage, has reportedly secured up to $300 million in insurance for AI-related risks, according to people familiar with the matter. Another source, however, disputed the figure, saying the coverage was significantly lower — a fraction of what would be needed to defend against a wave of potential lawsuits.
Aon declined to discuss individual clients, but Kevin Kalinich, the firm’s global head of cyber risk, said insurers lack sufficient capacity to cover the magnitude of potential claims facing AI providers. “What they can’t afford to pay is if an AI provider makes a mistake that ends up as a systemic, correlated, aggregated risk,” Kalinich told the Financial Times.
The insurance industry’s reluctance reflects both the novelty of AI-related liabilities and the potential scale of payouts, particularly as “nuclear verdicts” — massive damages awarded by U.S. juries — have become increasingly common against large technology firms. OpenAI currently faces multiple lawsuits, including a high-profile copyright case brought by The New York Times and several authors who allege their work was used to train models without permission. The company has also been sued for wrongful death by the parents of a teenager who died by suicide after using ChatGPT to discuss methods of self-harm.
According to people familiar with the matter, OpenAI has explored “self-insurance” — setting aside investor capital to cover potential liabilities — as it seeks more sustainable financial protection. One person said the company has considered establishing a “captive” insurance structure, a ringfenced entity often used by large corporations such as Microsoft, Meta, and Google to manage emerging risks like cyberattacks and social media liability. While captives can offer flexibility, experts caution that a major payout could deplete such funds, leaving companies exposed.
OpenAI, which has raised nearly $60 billion to date, said it maintains insurance coverage and continues to assess different insurance structures as it scales, but does not currently operate a captive and declined to comment on future plans.
Anthropic, meanwhile, has reportedly agreed to pay $1.5 billion to settle a class-action lawsuit brought by authors who accused the company of using pirated books to train its AI models. In court filings reviewed by the Financial Times, Anthropic’s legal team warned that the case carried “potentially business-threatening statutory damages” for smaller AI developers using the same datasets.
The company, which has raised more than $30 billion, is partially using its own funds to finance the settlement, a person with knowledge of the deal said. Anthropic declined to comment.
As insurers hesitate to assume the full weight of AI-related risks, both OpenAI and Anthropic are being forced to rely more heavily on investor backing and internal reserves — a sign that, even for the industry’s most prominent players, the cost of pioneering artificial intelligence could increasingly fall on their own balance sheets.
