A GlobalData survey found that SMEs in the US place the most value on protection against AI-related lawsuits. Meanwhile, Cowbell will launch a standalone AI liability policy for US SMEs next year, covering losses caused by AI tools that operate outside their intended boundaries.

GlobalData’s 2025 SME Survey found that 42.6% of US SMEs ranked coverage for legal fees and damages from AI-related lawsuits as the most valuable feature of a specific AI liability policy. This is well above the 36.3% average across the 14 countries surveyed, and it was the most popular first choice in both the US and globally. In the US, coverage for the costs of fixing or retraining a faulty AI model followed (32.9%), while regulatory fines and penalties were least likely to be ranked first (28.3%). The findings suggest US businesses see litigation as the most pressing financial risk of AI adoption, reflecting the country’s highly litigious business environment.

Meanwhile, Cowbell will introduce Cowbell AI Liability Insurance from next year, distributed exclusively through wholesale brokers on a non-admitted basis. The policy is aimed at businesses using AI tools across customer support, payment processing, and core operations. Claims are triggered by an “AI incident,” which Cowbell defines as an AI tool breaching its authorised boundaries, operating without the necessary clearance, exposing data without authorisation, producing inaccurate or hazardous output, or evading established protocols. Qualifying events also include model drift, algorithmic discrimination, and adversarial manipulation.

The policy appears closely aligned with the priorities identified in GlobalData’s survey. On the third-party side, it may cover defence costs, damages, or settlements linked to algorithmic bias, copyright or trademark infringement, and bodily injury or property damage resulting from autonomous AI actions or reliance on AI output. First-party cover includes the costs of isolating affected tools, investigating incidents, and recalling or correcting defective outputs, as well as direct financial losses from runaway tasks and “LLMjacking” incidents involving stolen API keys or credentials. Where permitted by law, it may also extend to regulatory fines and penalties. Underwriting will draw on three Cowbell Factors measuring AI observability, autonomy, and governance.

Overall, Cowbell’s launch reflects a growing recognition that AI risks are not fully addressed by existing cyber, technology errors and omissions, or general liability policies. By placing litigation protection at the core of the product, the policy targets the exposure US SMEs value most. However, demand will depend on how clearly insurers can define an AI incident and communicate what is and is not covered, particularly as cover remains subject to policy terms, limits, and exclusions. Insurers entering this space will need to offer straightforward wording and work closely with brokers to show SMEs where AI liability cover sits alongside their existing protection. Those that do so early will be well placed as AI adoption among smaller businesses accelerates.