For companies that build and deploy AI
Mayflower Specialty writes affirmative AI coverage for D&O, EPL and E&O, plus a DIC excess layer, for claims that arise from AI hallucinations, AI-driven cyber attacks and autonomous AI decisions. Policies are written on A- (Excellent) AM Best rated paper backed by global reinsurers.
Mayflower Specialty provides AI liability insurance for companies that build and deploy AI. It writes affirmative coverage for AI claims against directors and officers, employers and professional service providers, plus an AI DIC Excess layer over an existing program, on A- (Excellent) AM Best rated paper. Coverage can be bought directly from Mayflower, through a broker or embedded through partners that offer it alongside their own products.
Recognition
Selected for the FinTech Innovation Lab New York 2026 class
A 12-week program established by Accenture and the Partnership Fund for New York City.
Read about the selectionThe exposure
Public companies now disclose AI as a material risk, AI-related securities class actions are rising and courts are holding companies and lawyers to what their AI produces. The question for most companies is whether their insurance will answer an AI claim when one arrives.
72%
of S&P 500 companies flagged AI as a material risk in 10-K filings available by August 2025, up from 12% in 2023
Source:The Conference Board, October 2025 (opens in a new tab)
15
AI-related securities class actions were filed in the first half of 2026, putting such filings on pace to nearly double the 2025 total
2,100+
court decisions worldwide have dealt with hallucinated AI content submitted in legal proceedings, as of October 2026
Source:AI Hallucination Cases database, D. Charlotin (opens in a new tab)
Silent AI is the uncertainty that arises when an insurance policy neither covers nor excludes losses caused by artificial intelligence, which leaves the question of whether it responds to an AI claim to be argued after the loss.
Insurers have started to settle that argument with exclusions. W. R. Berkley has introduced an “absolute” AI exclusion for D&O, E&O and fiduciary liability policies that reaches any claim “arising out of” the “use, deployment, or development of Artificial Intelligence by any person or entity” Hunton (opens in a new tab), and Verisk's generative AI exclusions for general liability policies took effect in January 2026 Business Insurance (opens in a new tab).
An AI exclusion added at renewal can remove cover a company has relied on for years. AI DIC Excess is written to add affirmative AI coverage over the program you already carry, where those policies are silent on AI or exclude it, so you can keep your current carriers. It is a good idea to read each renewal for new AI wording before deciding whether that layer is needed.
Coverage
Mayflower's program is written around the ways AI fails in practice, from hallucinated outputs and AI agents that act beyond their instructions to biased automated decisions and AI-driven cyber attacks such as prompt injection and deepfake fraud. Each module is written for the claims those failures bring against a different part of the business, subject to the policy terms, so the right combination depends on where a company's AI touches its shareholders, its workforce and its clients.
Who brings the claim: Shareholders and regulators
Written for claims against directors and officers arising from the company's use, oversight or disclosure of AI, such as securities and derivative claims alleging AI-washing or failed oversight.
Who brings the claim: Job applicants and employees
Written for employment claims arising from AI used in hiring, promotion, discipline and other workforce decisions, such as discrimination claims over automated screening.
Who brings the claim: Clients and other third parties
Written for claims by clients and third parties who suffer a loss from AI-enabled products or services, such as an inaccurate or hallucinated output.
Who brings the claim: Any of the above, when your current policies exclude AI or are silent on it
A difference-in-conditions layer over an existing D&O, EPL and E&O program that adds affirmative AI coverage where those policies are silent on AI or exclude it.
What any policy covers depends on its wording. Module pages set out what each one is written for.
Courts and regulators
These public matters show the kinds of liability AI already creates and where in a company each one lands. Each is summarized from the sources linked, and whether a policy would respond to a similar claim depends on its terms and the facts.
B.C. Civil Resolution Tribunal, February 14th 2024
Inaccurate chatbot answer
Air Canada's website chatbot gave a customer wrong advice about bereavement fares. The tribunal held the airline liable for negligent misrepresentation and ordered it to pay C$812.02, rejecting the argument that the chatbot was a separate entity responsible for its own actions; in the tribunal's words, the chatbot “is still just a part of Air Canada's website.”
The kind of liability
A company can be held to what its AI tells customers. The award was small, but the reasoning could reach any business whose AI gives customers or clients wrong information that they rely on.
Written for this kind of claim: AI-E&O
U.S. District Court, N.D. California, Filed February 2023, pending
Biased automated decisions
A job applicant alleges that AI screening software Workday supplies to employers rejected his applications because of his race, age and disability. In July 2024 the court allowed disparate impact claims to proceed, and in May 2025 it granted preliminary certification of a nationwide collective of applicants aged 40 and over. The claims are allegations, and the case was in discovery as of August 2026.
The kind of liability
Automated hiring decisions can produce discrimination claims at collective scale, and employers are being sued as well as vendors: a proposed class action filed against Sirius XM in August 2025 alleged that the AI screening tool it used relied on proxies for race, although the court dismissed that complaint without prejudice in September 2026.
Written for this kind of claim: AI-EPL
U.S. Securities and Exchange Commission, March 2024 and January 2025
AI-washing
In March 2024 the SEC settled charges against two investment advisers, Delphia and Global Predictions, for false and misleading statements about their use of AI, with $400,000 in combined civil penalties. In January 2025 it settled charges that Presto Automation made materially false statements about its AI drive-thru ordering product, in what the D&O Diary described as apparently the SEC's first AI-washing action against a public reporting company.
The kind of liability
Statements about AI are disclosures, and overstating them can lead to regulatory actions and shareholder suits that name directors and officers, which is the territory D&O insurance exists to defend.
Written for this kind of claim: AI-D&O
Reported by Fortune, July 2025
Agentic AI action
During a code freeze, an AI agent on Replit's platform ran commands that deleted a customer's live production database, despite instructions not to act without human approval. The data was later recovered, and Replit's chief executive called the failure “unacceptable” and announced new safeguards.
The kind of liability
No lawsuit was reported, but when an AI agent acting for a vendor or service provider damages a client's systems or data, the client's loss can turn into a professional liability claim against that provider.
Written for this kind of claim: AI-E&O
How it works
The controls a company already runs, such as an AI inventory, an oversight body and bias testing, can count in its favor when Mayflower underwrites the application, so it is worth documenting them before you apply.
Start the online application, which mirrors Mayflower's AI Liability Supplemental Application and saves your progress, or work from the PDF version. A broker can also apply on your behalf.
Download the PDFUnderwriters read the application section by section, from the AI systems you run and how they are governed to data, monitoring, incident response, regulation and claims history.
Mayflower's underwriting team reviews your submission and responds within 48 hours. Completing the application does not bind coverage, so there is time to review any terms before deciding to buy.
Terms come back to whoever submitted the application, you or your broker, and nothing is bound until you decide to buy. The policy is written to sit alongside the rest of your insurance program.
Talk to usUnderwritten on your governance
After the applicant details, the application asks about seven areas of AI risk and governance. Each one tells underwriting something about the size of the exposure or the controls that contain it.
The number of AI systems, how autonomous they are and whose decisions they affect describe the exposure underwriting assesses.
Underwriting looks at how AI is overseen, documented and approved before a system reaches production.
The data a model learns from and acts on can give rise to bias, privacy and intellectual property claims, so underwriting asks how it is controlled.
Monitoring and fallback plans show how quickly a failing system would be noticed and contained.
Underwriting assesses how the company would detect, contain and learn from an AI incident.
The laws and regulators that apply to your AI systems shape the regulatory exposure underwriting assesses.
Past claims, known circumstances and prior declinations are part of underwriting any liability policy.
The application also lists recommended documents, such as board or committee minutes on AI risk oversight, bias audit results and an AI incident response plan, which may improve terms.
Why Mayflower
These principles shape how Mayflower writes AI liability coverage, from the policy form and the underwriting to the paper and the claims handling behind each policy.
Many policies cover AI only because they say nothing about it, and that cover can disappear when an exclusion arrives at renewal. Mayflower's forms name the AI exposure and cover it deliberately, either as a modular primary policy or as a difference-in-conditions layer over the program you already carry.
Terms reflect how a company actually runs AI: how decisions are made, monitored, documented and escalated. Strong governance can earn broader terms and better pricing, because underwriting gives credit for the controls a company has already built.
AI-D&O is written for the board's duty to oversee the AI systems a company relies on, AI-EPL for automated hiring and workforce tools that are already drawing collective actions, and AI-E&O for AI-enabled products and services that cause a client a loss. AI DIC Excess adds affirmative AI coverage over an existing D&O, EPL and E&O program.
Coverage is written on A- (Excellent) AM Best rated paper, backed by some of the world's largest reinsurers, and claims are handled by a specialist claims administrator. AM Best's Excellent category, which covers its A and A- ratings, is the second highest of its Financial Strength Rating categories and is assigned to insurers that in AM Best's opinion have an excellent ability to meet their ongoing insurance obligations.
AI liability and cyber
Cyber insurance usually responds to an AI claim only when it involves a security or privacy event, such as a breach, ransomware or a network outage. Many AI claims involve no breach at all, for example a chatbot giving a customer wrong advice or a hiring tool screening out older applicants, so they fall to D&O, EPL and E&O policies instead, which may be silent on AI or exclude it.
The two kinds of policy answer different triggers, claimants and defenses, so they are designed to sit beside each other. The sensible course is to keep cyber cover for security events and to check separately that the D&O, EPL and E&O program will answer an AI claim.
| Question | Cyber insurance | AI liability insurance |
|---|---|---|
| What sets it off | A security or privacy event, such as unauthorized access, a data breach or ransomware | A decision or output from an AI system that harms a person or a business, often with no breach at all |
| Who brings the claim | The company itself for its response costs, and people whose data was exposed | Customers and clients, job applicants and employees, shareholders and regulators |
| A typical example | Ransomware locks a company's systems and customer records are stolen | A chatbot gives a customer wrong fare information and the company is held to it, as in Moffatt v. Air Canada |
| Where it is insured | A cyber policy | D&O, EPL and E&O policies, with AI-D&O, AI-EPL, AI-E&O and AI DIC Excess written for the gaps |
AI-driven attacks sit on the boundary between the two: a prompt injection that leaks customer data looks like a cyber event, while one that leads an AI system to give harmful advice or take a harmful action can become a liability claim, which is why the AI wording in both policies matters.
Coverage gap check
The coverage gap check asks how your company uses AI and which policies it holds, then shows where an AI claim may fall outside your current program and which Mayflower module is written for that gap.
The check gives general information and does not review your policy wording.
| Scenario | A standard policy | Written for it |
|---|---|---|
| A customer relies on a chatbot's wrong answer and sues | An E&O policy may be silent on AI or exclude it, depending on the wording | AI-E&O |
| Applicants allege that an AI screening tool rejected them because of their age | An EPL policy may respond, unless an AI exclusion applies | AI-EPL |
| Shareholders allege that the company overstated its use of AI | A D&O policy may respond, unless an AI exclusion was added at renewal | AI-D&OAI DIC Excess |
These scenarios are illustrative, and whether a policy responds depends on its wording and the facts of a claim.
Guides
Plain answers to the questions risk managers, general counsel and brokers ask about AI liability, with the cases and sources behind them.
Underwriters price AI liability on three things: how much harm the company's AI could do to people and their finances, how well the company governs and monitors it, and its claims history. Preparing an inventory of AI systems, a written governance policy and evidence of testing before applying gives the underwriter what it needs and can improve terms.
12 min read · October 5th 2026Silent AI is the uncertainty that arises when an insurance policy neither covers nor excludes losses caused by artificial intelligence. Insurers are now resolving it with AI exclusions, including ISO general liability endorsements effective January 2026 and at least one far broader exclusion in D&O and E&O forms, so each renewal should be read for new AI wording and affirmative cover bought where it is removed.
11 min read · October 5th 2026Employers that use AI to screen, rank or evaluate people remain responsible for the discrimination those tools cause, and a federal court has allowed claims against the software vendor as well. Rules in New York City, Illinois, California and Colorado add notice, audit and record-keeping duties, and an EPL policy that is silent on AI or excludes it may not respond to these claims.
12 min read · October 5th 2026Questions
These are short answers for buyers and brokers, and each topic has a fuller treatment on the coverage pages and in the guides.
More answers in the FAQTwo ways to start
The online application asks the same questions as Mayflower's supplemental application and saves your progress as you go. If you would rather talk first, a short message reaches the team.
Completing the application does not bind coverage, and underwriting responds within 48 hours.
