Hunton Insurance Recovery Blog

The insurance market’s approach to artificial intelligence is evolving almost as quickly as the technology itself.

Last year, we highlighted the emergence of affirmative AI insurance, offering clear coverage for AI-related risks rather than relying on protections under legacy policies. At the same time, insurers were moving in the opposite direction in traditional policies, introducing “Absolute” AI exclusions for certain D&O, E&O, and fiduciary liability coverages.

Another development suggests the affirmative side of that market is continuing to mature.

Cowbell has announced a standalone AI Liability Insurance policy for US small and medium-sized businesses using AI agents. Scheduled to become available January 1, 2027, the coverage centers on a defined “AI Incident” and is designed to address both first-party losses and third-party claims when AI systems exceed their authority, act without required approval, expose data, or generate inaccurate or unsafe outputs.

Potential coverage includes incident-response and output-correction costs, misdirected transactions, runaway AI usage charges, and “LLMjacking,” as well as claims alleging algorithmic discrimination, intellectual property infringement, bodily injury, or property damage. This product supplements the insurer’s existing affirmative AI coverage, a Prime One cyber product for larger organizations.

For organizations using agentic AI, this should come as a welcome development as more and more AI agents make headlines for the wrong reasons—from submitting false police tips and inserting malicious code to deleting company databases and breaking into real companies during security testing. As the insurance market continues to develop, AI risk is increasingly being addressed expressly rather than left to traditional policy language. Businesses assessing their insurance programs against those risks should evaluate both sides of the equation: where their existing policies may be narrowing and whether emerging AI-specific products could fill the resulting gaps.