What it is
As businesses adopt AI at scale, the liability exposure is falling through the gaps of traditional insurance programs. Professional liability, E&O, D&O, EPL, CGL, and cyber policies were not designed to address human oversight failures in AI-assisted work, or harm caused by autonomous systems. Axiom's AI liability policy was.
Coverage Build Across 10 Insuring Agreements
We provide affirmative third-party coverage for claims arising from the usage and deployment of AI. Each of our 10 IAs is bindable on its own and shared a single aggregate limit. Buy what you need, leave what you don't.
Coverage | Description |
|---|---|
Generative AI Errors & Omissions | Covers third-party financial loss from wrong content an AI produces - a hallucination, error, or misstatement - that someone relies on. Responds whether a person relayed the output or an agent sent it autonomously. |
Autonomous Action Liability | Covers loss when an AI agent takes an action outside its authorized scope, or inside scope with a containment failure. This is the coverage for when an agent acts rather than just producing content. |
AI Discrimination & Employment Practices | Covers discrimination, wrongful termination, retaliation, and failure-to-hire claims where AI participated in the decision, plus consumer-side discrimination in lending, housing, and service. |
AI Regulatory Proceedings | Covers defense costs and insurable fines when a regulator alleges your AI deployment violated an AI-specific law such as the EU AI Act, Colorado AI Act, or NYC Local Law 144. |
AI Bodily Injury & Property Damage | Covers physical injury to people or damage to property caused by reliance on AI output or by an agent's action. Sits in the gap between general liability and E&O. Every submission is reviewed and priced by an Axiom underwriter. |
AI Intellectual Property Liability | Covers defamation, copyright or trademark infringement, false endorsement, and right-of-publicity violations in AI-generated content, plus infringement arising from the insured's own fine-tuning or retrieval-augmented-generation practices. Written excess and difference-in-conditions to vendor indemnities and to any responding media or general liability policy. |
Wrongful Data Disclosure | Covers loss arising when the insured's own declared AI discloses protected information through its output or action, without requiring any security failure. Written excess to any cyber policy that might also respond. |
AI System Manipulation Liability | Covers liability when a third party manipulates your AI through prompt injection, jailbreak, or data poisoning using its intended interface - attacks cyber policies don't treat as a breach. |
AI Management & Fiduciary Liability | Covers directors, officers, and plan fiduciaries facing claims over AI oversight, AI-related disclosures, or AI-driven benefits administration where the underlying D&O or fiduciary policy excludes AI. |
Contingent Vendor AI Liability | Covers the insured's loss and third-party liability arising from a declared AI vendor's failure - a vendor-side breach, an abrupt shutdown, a material undisclosed model change, or a licensing collapse - beyond service-level and contractual recovery. Sublimited and subject to Axiom underwriter referral. |
What Triggers a Claim
The policy operates on a claims-made basis. A claim is triggered when a third party suffers demonstrable harm resulting from your firm's inadequate oversight of an AI-assisted process or output. The policy insures the oversight failure - not the performance of the underlying AI model.
How underwriting works
Our underwriting follows a defined five-step path. We start with the application, perform a non-intrusive external scan, assign a risk class from inherent hazard and declared controls, bind coverage, and deploy Raptor at no additional cost. The process is the same for every account, and every decision is documented.
What makes it different
Most policies address AI in one of two ways: they exclude it explicitly, or they are silent on it entirely - leaving coverage to interpretation at the worst possible moment. Our AI liability policy is affirmative. Every material term, trigger, and exclusion is defined. You know exactly what you have before you need it.
Hypothetical Claim Scenarios
Industry | AI System | Scenario | Coverage |
|---|---|---|---|
Legal Services | LLM Research Assistant | A content-summary AI misstates service terms, and the firm misses a contractual obligation as a result. | Gen-AI E&O |
Enterprise Support | Customer-Facing Chatbot | A support chatbot gives an enterprise client incorrect integration steps, causing downtime and lost revenue. | Gen-AI E&O |
Financial Services | Agentic Procurement AI | A procurement agent with purchasing authority autonomously places an order with a sanctioned supplier. | Autonomous Action Liability |
Professional Services | Agentic Ops Assistant | An operations agent with production access deletes client deliverables while "resolving" a storage alert. | Autonomous Action Liability |
Staffing / HR | LLM Resume Screener | An AI screening tool produces a disparate-impact class action and penalties under NYC Local Law 144 | Discrimination & EPL |
Banking | LLM Service Tier | An AI customer-service tier offers materially worse terms to customers correlated with protected class. | Discrimination & EPL |
E-Commerce | Retail Chatbot | A support chatbot reveals another customer's order history, bank details, and address in a reply. | A Wrongful Data Disclosure |
Professional Services | Meeting Transcription AI | A transcription bot emails one client's summary, including confidential deal terms, to a different client. | Wrongful Data Disclosure |
Telecoms | Customer Service Chatbot | A state regulator opens an investigation alleging the company's AI chatbot violated state AI transparency rules | AI Regulatory Proceedings |
Insurance | AI Claims Triage System | A state attorney general investigating the insurer's AI triage under a new state AI statue. | AI Regulatory Proceedings |
Real Estate | Agentic Building Controls AI | An AI assistant sets a building's HVAC to the wrong temperature, freezing and bursting pipes in a tenant's unit. | Bodily Injury & Property Damage |
Healthcare / wellness | LLM Wellness Assistant | A wellness AI gives unsafe recovery guidance that allegedly causes physical injury to a user. | Bodily Injury & Property Damage |
Marketing | Generative Contant AI | An AI campaign uses a tagline that turns out to be a competitor's registered trademark, triggering a complaint. | Intellectual Property |
Media | Image Generation AI | A marketing image materially replicates a photographer's registered work after guardrails were disabled. | Intellectual Property |
Fintech | Agentic AP Assistant | A prompt-injected email tricks an accounts-payable agent into approving a chain of fraudulent invoices. | System Manipulation |
Consumer Platform | Customer Facing Chatbot | A jailbroken chatbot emits harassing content at a user, who then sues the company. | System Manipulation |
Public Company | Enterprise-wide AI Program | A securities class action alleges the company overstates its AI capabilities in investor materials. | Management & Fiduciary |
Employer / Plan Sponsor | AI Benefits Administration | An AI-driven retirement plan menu concentrates participants in underperforming funds, prompting an ERISA suit. | Management & Fiduciary |
Consulting | Third-party Legal AI Vendor | A declared legal-AI vendor suffers a breach exposing the insured's client files held on the vendor's systems. | Contingent Vendor |
Professional Services | Third-party Agentic Vendor | A vendor abruptly shuts down mid-engagement, causing client facing losses beyond any service credit. | Contingent Vendor |