Artificial intelligence and autonomous technologies are rapidly changing liability risks, with traditional insurance policies struggling to cover these new exposures. Recent data shows AI-related lawsuits in the US have risen sharply, including cases of algorithmic bias[1], copyright issues, AI hallucinations[2], and even actual physical harm—such as false medical advice from chatbots and accidents caused by autonomous vehicles. Insurers are responding with new exclusions and specialized products, and policyholders must review their coverage, consult insurers about dedicated AI solutions, and negotiate endorsements for emerging risks.
The Situation and Emerging Issues
Current insurance market conditions show persistent hardening in casualty lines. Umbrella and excess liability policy premiums are experiencing continued upward pressure, with liability rates rising for many insureds in the first quarter of 2026. This trend is especially pronounced for risks associated with technology-intensive and high-hazard industries. Insurers are responding by reducing primary policy limits—often capping coverage at approximately $5–10 million for more challenging classes of risk—while increasing attachment points to better manage the impact of social inflation and large court verdicts.
Traditional Commercial General Liability (CGL) policies, which typically serve as the foundation for umbrella coverage, are increasingly providing only fragmented protection against AI-related risks. These include model drift, where AI systems unexpectedly change performance over time—such as making inaccurate predictions following software updates—and third-party tool liabilities, which arise when businesses face legal action due to errors from outsourced AI analytics platforms or other external AI tools.
The unpredictable and autonomous nature of AI systems is at the forefront of emerging insurance concerns. "Agentic AI," systems capable of performing multi-step tasks such as signing contracts or operating machinery—raise issues related to agency law and vicarious liability. AI hallucinations have already been the subject of defamation and personal injury lawsuits, and accidents involving autonomous vehicles are increasing bodily injury claims due to algorithmic decision-making errors. State-level AI legislation introduced or expanded in 2025–2026 has created new private rights of action for algorithmic discrimination and deepfakes, which may affect umbrella coverage for businesses working across multiple states. Additionally, insurers' use of AI in underwriting and claims handling add risk as well, including potential bad faith allegations if AI-driven decisions result in the denial of valid claims.
The Lawsuits
Key lawsuits illustrate these new frontiers in liability. By late 2025, cumulative GenAI (generative artificial intelligence) cases have exceeded 700 in number, spanning issues such as copyright disputes, algorithmic bias, and bodily injury claims. In A.F. et al. v. Character Technologies, Inc., plaintiffs alleged psychological and emotional harm resulting from an AI chatbot's interactions, raising questions about whether such claims trigger standard bodily injury coverage under CGL (Commercial General Liability) excess/umbrella layers.
The court's interpretation of bodily injury in such context could redefine how insurers assess AI-related risks, potentially influencing future policy language and claims handling. Securities lawsuits related to overstated AI capabilities, including those focused on autonomous vehicle disclosures—have implications for Directors & Officers (D&O) excess towers, where interrelated claims may aggregate to single policy limits.
Coverage disputes are emerging over whether AI-related harms constitute "professional services" or fall under silent cyber-physical convergence, often resulting in disagreements about follow-form language (contract terms in excess policies that mirror primary policies) in excess policies. Continuing to monitor real-world case examples and evolving legal trends is critical for stakeholders seeking to navigate the shifting landscape of GenAI liability.
Insurers Push Back
New ISO endorsements, specifically CG 40 47 and CG 40 48, effective January 2026, allow insurers to exclude coverage for bodily injury, property damage, or personal and advertising injury resulting from generative AI outputs—content produced by artificial intelligence models. As a result, policyholders may need to seek supplemental coverage or face gaps in protection if their operations rely on generative AI technologies. These and other new AI exclusions are becoming widespread, prompting many insureds to seek more affirmative AI coverage to bridge the resulting coverage gaps.
On the Horizon
Looking ahead in 2026, further guidance from the NAIC on AI transparency is expected, along with potential tensions between federal and state regulatory approaches. Tort reforms in states such as Florida and Georgia—which cap non-economic damages and increase disclosure requirements for litigation funding—may help moderate overall social inflation, but AI-specific liabilities could counteract any softening effects in technology sectors. Insurers are likely to refine absolute AI exclusions, while specialty markets, including Lloyd's syndicates, rush dedicated AI liability products to market. Quota-share arrangements in excess insurance towers are anticipated to become more common as carriers seek to spread the risk associated with undefined AI severity.
Conclusion
AI and autonomous technologies are ushering in a new era for umbrella coverage in 2026. While market hardening and emerging exclusions present significant challenges, diligent policy reviews and strategic layering of coverage can help policyholders secure the protection they need amid increasing claim frequency and severity.
To minimize coverage issues in layered insurance programs, policyholders should negotiate robust follow-form endorsements that clearly align excess terms with any AI-related modifications in primary policies, thereby preventing gaps created by new ISO exclusions. In quota-share policies, it is vital to establish clear protocols for settlement authority and allocation of defense costs to avoid bad faith claims, especially when carriers differ on whether an AI hallucination is excluded. Evaluating the financial strength of insurers remains essential, as ambiguities in AI wordings may result in disputes like those seen in early cyber coverage cases. For high-AI exposures, policyholders might consider captive insurance arrangements or parametric solutions[3]. Make sure your broker is highly versed in the identification and treatment of AI liability risks, interview alternative qualified brokers/consultants as needed.
___________________________
[1] Algorithmic bias refers to the tendency of artificial intelligence systems or algorithms to produce results that are systematically prejudiced due to erroneous assumptions, flawed data, or design choices. This bias can lead to unfair discrimination against certain groups or individuals, often manifesting in decisions about hiring, lending, healthcare, or law enforcement.
[2] Confident but false, inaccurate, or nonsensical information generated by artificial intelligence, particularly large language models (LLMs). These outputs appear grammatically correct and realistic but are fabrications not supported by training data, often caused by inadequate data, overfitting, or flawed logic
[3] Non-traditional insurance model that pays out a fixed amount automatically when a specific, predefined trigger event occurs
(Photo credit: ZinetroN/Adobe Stock)
This article originally appeared on FC&S Expert Coverage Interpretation and may not be reprinted.
Related:
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.