Anthropic IPO exposes major AI insurance coverage gaps and liability risks

Anthropic’s IPO filing has brought renewed attention to a question confronting the insurance industry: whether insurers can provide sufficient coverage for the financial, legal and physical consequences of artificial intelligence failures.

The AI developer’s S-1 registration statement identifies risks ranging from intellectual property disputes and security breaches to data loss, autonomous AI agents, climate-related exposures and potentially catastrophic harm. The company acknowledges that its existing insurance arrangements may be insufficient to address the scale of some incidents.

• Anthropic warns that catastrophic AI incidents could generate liabilities beyond the capacity of individual insurers or the wider insurance market, potentially leaving substantial losses uninsured.
• Cyber insurance currently covers some AI-related financial and security losses, but bodily injury, physical damage and unintended AI behavior remain difficult to insure.
• Zurich, Marsh and other insurance specialists expect AI underwriting, pricing and exclusions to evolve, although they disagree on whether AI needs dedicated policies or coverage under existing insurance products.

“Because our models may behave in ways that no user or developer intended, claims arising from such actions could result in substantial liability,” Anthropic stated in its filing.

According to CNBC, the company also warned that a single event, or multiple related incidents, could produce losses disproportionate to the insurance protection available.

Anthropic’s financial resources do not necessarily resolve that problem. The company said the potential consequences of a catastrophic AI incident could exceed the underwriting capacity or expertise of any individual insurer and possibly the insurance industry as a whole.

The market for AI-specific insurance remains relatively undeveloped, according to the filing, and some liabilities may be impossible to insure at commercially reasonable prices.

That creates the possibility of substantial uninsured losses even for companies with significant cash reserves.

The concern extends beyond developers of advanced AI models. Businesses are incorporating AI into customer service, recruitment, financial transactions and operational decisions, sometimes without fully understanding how existing liability policies respond when the technology causes harm.

Recent incidents involving AI chatbots behaving unexpectedly have intensified scrutiny. Some businesses have restricted access to autonomous AI agents, while lawsuits have alleged that chatbot interactions contributed to deaths, including a case reported during the week of Anthropic’s filing.

These developments are forcing insurers to examine whether established policy language adequately addresses the consequences of AI-generated decisions and actions.

AI liability claims raise questions about insurance capacity

Alissa Lugo, senior director analyst in Gartner’s legal and compliance practice, said her research indicates that claims involving deaths in which AI played some role could exceed 2,000 worldwide by the end of 2026.

The projection includes cases where AI contributed to a fatality, rather than necessarily being established as the sole cause. Such claims can present difficult questions about responsibility, causation and the applicable insurance coverage.

Lugo said insurers are increasingly introducing AI exclusions, while courts are likely to help determine how existing policy language applies to emerging claims.

Bodily injury represents a particularly difficult area because cyber insurance, currently one of the main sources of protection for AI-related losses, generally concentrates on digital and financial harm.

Dennis Hendrickson of Hendrickson Insurance, an independent brokerage in Sarasota, Florida, said cyber policies can respond when AI contributes to data breaches, privacy violations, system outages or fraudulent transfers involving deepfakes.

Some cyber insurers have expanded their policy wording to include AI-driven security incidents. Those adjustments can provide protection against familiar cyber threats involving new technology, but they do not necessarily address physical injuries.

The absence of bodily injury and physical damage protection as a substantial coverage gap that insurers will need to address.

Zurich identifies gaps in AI insurance coverage

Kara Higginbotham, head of professional liability and cyber for U.S. national accounts at Zurich North America, said AI is introducing exposures across several insurance classes, although insurers are not approaching those risks consistently.

Zurich has approved certain AI tools for employee use while restricting others. Higginbotham said similar controls may eventually become standard requirements when insurers assess companies seeking AI-related protection.

For now, bodily injury and physical harm resulting from AI incidents are generally not covered automatically.

Insurers have limited historical claims experience for some AI-related incidents, making it difficult to estimate potential losses or establish adequate premiums.

General liability policies could sometimes be extended to address certain exposures, but insurers would need to assess the circumstances and potential severity of claims.

Companies cannot always identify which employees are using AI, what decisions the tools influence or whether third-party software contains AI capabilities.

Henrietta Akpata, executive fellow and visiting professional lecturer of management and AI at Marist University, said insurers increasingly expect businesses to demonstrate how they supervise AI use. “Insurers are saying, ‘We can’t cover what you can’t show us.’ Most companies can’t say where their employees are using AI tools. That’s the first thing to fix,” Akpata said.

A small business could face liability if its website chatbot gives customers inaccurate information. An employer using AI to screen job applications could encounter claims arising from recruitment decisions.

Should AI risks be covered by existing insurance policies?

Insurance specialists disagree on whether AI liability will develop into a separate insurance market or remain within established commercial insurance products.

Coverage for AI-related financial losses to become relatively routine, with pricing and underwriting approaches similar to cyber insurance. Physical injury will be considerably harder to insure.

Some businesses operating close to healthcare may find suitable coverage unavailable altogether, according to insurance specialists.

Underwriting guidelines, premiums and other terms may need adjustment, but she questioned the value of treating AI as an entirely separate insurance category.

A privacy breach involving AI remains a privacy breach, while a security incident facilitated by AI remains a security incident. The underlying cause of loss may therefore be more relevant to insurance coverage than whether AI was involved.

Managing potentially overlapping policies could create additional complexity for customers without necessarily improving protection.

Kim said the insurance industry has not yet identified an entirely new category of loss created by AI, although she acknowledged that this could change.

Cyber insurance offers a precedent for AI underwriting

Several insurance specialists see similarities between current AI coverage uncertainty and the early development of cyber insurance.

Vikram Desai, global cyber strategy, risk and architecture lead at Accenture, said corporate discussions have moved from deploying AI to understanding the consequences when the technology fails.

Cyber exposures were initially addressed within broader liability insurance arrangements before insurers began identifying, excluding and subsequently offering coverage for those risks through more specialised products.

“Cyber risk sat silently inside general liability for years until insurers named it, excluded it, then sold it back. AI is on the same curve, but compressed,” Desai said. He identified undisclosed or poorly understood AI deployments as a particular concern.

Companies may face exclusions even when AI contributes to an incident through software supplied by another vendor.

The real exposure is the AI companies don’t know they have. Exclusions can bite when AI sits anywhere in the causal chain, including inside a vendor’s product.

Underwriters are already asking companies to identify where AI is deployed and explain who is responsible for overseeing its operation.

Those questions could become increasingly important when insurers determine whether to provide coverage, impose exclusions or require additional controls.

AI-related insurance claims are concentrated in intellectual property and privacy

Despite growing concern about catastrophic AI incidents, current claims activity has been concentrated in more established categories of liability.

Many AI-related claims observed by Zurich involve intellectual property disputes rather than large-scale cyberattacks or network failures.

These include allegations that AI developers collected copyrighted material to train their models without authorisation, along with claims concerning infringement arising from the use of protected content.

Zurich has also encountered claims involving misuse of data and privacy-related incidents.

AI’s role in cybercrime presents an additional complication because insurers cannot always determine whether attackers used the technology.

Social engineering attempts involving fraudulent emails or messages impersonating colleagues through platforms such as Microsoft Teams may employ AI-generated content without leaving clear evidence of its use.

The extent of AI’s involvement in individual incidents is not always evident, however.

She does not believe those exposures necessarily justify a separate AI insurance category because cyber policies already address the underlying security and privacy risks.

Insurers may eventually need to increase premiums if AI materially changes the frequency or severity of claims.

AI insurance exclusions and underwriting standards could expand

Zurich expects its underwriting practices to change as insurers accumulate experience with AI-related claims and better understand emerging threats.

The company is working with law firms and technology providers to improve its approach to assessing and managing these exposures. Rather than automatically excluding new risks, Zurich intends to help clients establish appropriate safeguards.

Companies that fail to implement baseline controls or refuse to improve their risk management practices could eventually face exclusions.

Insurers may initially introduce new products before discovering that actual losses exceed their expectations, potentially leading to tighter renewal terms or decisions not to renew coverage.

Independent brokers could then face complaints from clients who believed they had adequate insurance but subsequently discover that particular AI-related losses are excluded.

Insurance requirements to influence corporate AI governance before governments establish detailed regulatory standards.

For Anthropic and other AI developers, the unresolved issue is whether insurance markets will eventually offer coverage proportionate to the potential consequences of advanced AI failures.

For insurers, the immediate question is how to assess exposures that are becoming embedded in ordinary commercial activities, often without sufficient information about deployment, oversight or potential losses.

The decisions insurers make about policy wording, underwriting controls and exclusions will determine how much of that liability businesses can transfer and how much they must retain themselves.