Updated 17 August 2026 for the EU AI Omnibus, in force since 27 July 2026.

The AI Omnibus entered into force on 27 July 2026 and moved the dates this article was originally written against. High-risk obligations for Annex III systems now apply from 2 December 2027, and for Annex I systems embedded in regulated products from 2 August 2028. The transparency and governance rules, Article 50 among them, applied from 2 August 2026, and the AI Office and the national authorities have held implementation, supervision and enforcement responsibility since that date. Article 5 prohibitions, the Article 4 AI literacy duty and the GPAI obligations under Articles 53 and 55 are unchanged, as is the 9 December 2026 transposition deadline for the revised Product Liability Directive. Source: European Commission, regulatory framework for AI, checked 17 August 2026.

Underwriting does not follow the deadline. Specialist carrier capacity kept expanding through the whole period in which the delay was uncertain, and specialist underwriters price to operator practice rather than to regulatory dates. What the Omnibus in force changes for AI agent insurance sets out the consequences for cover.

In short
  • At least twelve carriers or programmes are writing meaningful AI liability limits as of April 2026, concentrated in the US specialty and Lloyd's markets. No European-domiciled primary AI insurer exists yet.
  • Per-risk capacity ranges from standard commercial limits for SME products (HSB) through USD 9.25 million per insured (Testudo, since 26 February 2026) and USD 25 million (Armilla) to bespoke enterprise programmes via Lloyd's. Layered programmes combining multiple carriers can plausibly reach USD 50 million or above, on this desk's arithmetic from published single-carrier limits.
  • The market is structurally bifurcated: specialist carriers are building affirmative AI products while mainstream generalists are reported to be filing exclusions, AIG, W.R. Berkley, Great American and Chubb among them. Buyers cannot assume their existing policies respond to AI losses.
  • SME buyers now have a purpose-built product: HSB AI Liability Insurance, launched March 2026, covering bodily injury, property damage, and advertising injury from AI use, distributed through carrier partners.
  • Enterprise buyers face a documentation-intensive underwriting process. Five factors shape how an underwriter reads the risk: certification posture, autonomy envelope, deployment scope, sector sensitivity, and claims history. Organisations without structured AI governance documentation leave the underwriter more to estimate, which can show up as more questions and more conditions.
  • Underwriting criteria are converging around four frameworks: NIST AI RMF, ISO/IEC 42001, AIUC-1, and sector-specific standards. Third-party validation against any of these frameworks consistently shortens the underwriting process.
  • No carrier publishes a pricing floor for meaningful mid-market limits (USD 5 million) in 2026. Premium varies with autonomy level, sector, and governance maturity, and no figure is stated here.

Section 1. Market Overview: The Bifurcation

The AI liability insurance market entered 2026 split in two directions simultaneously. One half of the market is building specialist coverage for AI-specific risks at speed. The other half is filing exclusions to remove AI risk from the policies that previously covered it by default.

This bifurcation was documented by S&P Global Market Intelligence in February 2026, which noted that "as insurers retreat from AI risk, specialist startups plan to fill the gap." Trade publications including Insurance Times, Reinsurance News, and Business Insurance tracked the acceleration through Q1 2026, as the ISO exclusion endorsements CG 40 47 and CG 40 48 took effect on 1 January 2026, AIG and W.R. Berkley filed exclusion language with state regulators, and Testudo, HSB, and Armilla simultaneously announced new affirmative products.

The European Insurance and Occupational Pensions Authority surveyed the sector in February 2026, finding that GenAI use in the European insurance sector is growing rapidly but that coverage for AI liability remains concentrated in US and London markets. European primary capacity is still described as "pending" across most registries.

The practical consequence for buyers is significant. A business that assumed its commercial general liability, professional indemnity, or management liability policy covered AI-related losses may now find that assumption is wrong. The reported January 2026 effective date on the ISO exclusion endorsements created a specific cliff: policies renewed on or after that date at participating carriers may exclude generative AI losses that earlier policy years would have covered.

The AI-specific insurance premium pool is at an early stage in 2026, and no market-size figure is stated here. The 2026 market represents the accumulation period: underwriting data is thin, adverse selection is a real concern, and carriers writing the first generation of AI policies are doing so with limited actuarial experience. This structural immaturity affects both pricing and policy terms. It also creates the opportunity: the carriers and brokers that build underwriting expertise now will hold structural advantage as the market matures.

Section 2. How to Read This Map

The carriers profiled in this map can be classified along four independent axes. Understanding the axes allows a buyer or broker to navigate the market without treating all AI coverage as interchangeable.

Axis 1: Monoline Specialty vs. Extension

Monoline specialty carriers write AI liability as their primary or sole product. Armilla and Testudo, both Coverholders at Lloyd's, are monoline specialists: their entire underwriting focus is AI risk. AIUC is a standards body, not an insurer. Extension carriers add AI coverage to an existing product line. Counterpart extended its professional liability suite to include affirmative AI coverage in November 2025. HSB added an AI liability product to its specialty equipment breakdown and technology insurance portfolio. Vouch offers AI coverage as part of a broader startup insurance package. The distinction matters because a monoline specialist develops deeper AI-specific actuarial data, while an extension carrier may have broader distribution but shallower AI underwriting expertise.

Axis 2: Performance triggers vs. indemnity

Performance-based products settle when a measurable performance trigger is breached, without a full claims adjustment process. Munich Re aiSure is the primary performance-based product in the market. It is not parametric: Munich Re does not describe it that way, and aiSure settles on measurable performance data after Munich Re technical due diligence rather than firing on an index. Indemnity products respond to actual third-party claims. Armilla, Testudo, Counterpart, HSB, Corgi, and Vouch are all indemnity-based. The distinction affects claims speed, basis risk, and what losses are actually covered. A performance-based product may pay when a model underperforms even if no third party has sued; an indemnity product pays when a third party makes a claim, even if the underlying model performed within specification.

Axis 3: SME vs. Enterprise

HSB is currently the only carrier with a product explicitly designed and marketed for small and medium-sized businesses. Vouch and Corgi serve the startup and growth-stage technology company segment. Buyers in the revenue band this desk treats as mid-market, roughly USD 10 million to USD 500 million, are served by Counterpart and Testudo. The band is our segmentation, not an eligibility rule published by either. Buyers above that band, or deploying AI at scale in regulated sectors, need the higher limits available from Armilla, Munich Re aiSure and bespoke Lloyd's programmes. Again, the threshold is this desk's framing rather than a published carrier appetite statement.

Axis 4: US-Domiciled vs. EU vs. Global

All specialist AI insurance products writing meaningful limits are currently US-domiciled or routed through the London market. No European-native primary AI insurer exists as of April 2026. European buyers can access coverage through the Lloyd's coverholders Armilla and Testudo, or through other Lloyd's capacity via wholesale brokers. The European coverage gap is the structural gap this publication tracks. Until a European-domiciled carrier builds a product aligned with EU AI Act and Product Liability Directive language, European buyers are exposed to policy language written for a different regulatory environment.

Section 3. The Specialist Carriers

The following profiles are drawn from publicly available carrier announcements, trade publication reporting, and regulatory filings. Capacity figures are market observations and may not reflect current available terms. Premium estimates are illustrative and subject to underwriting. Where verification was not possible, claims have been omitted.

Munich Re aiSure

Product name: aiSure
Parent / capital: Munich Re Group, one of the world's largest reinsurers. Münchener Rückversicherungs-Gesellschaft Aktiengesellschaft in München.
Launch: Mosaic Insurance partnership announced 26 February 2026, with Mosaic underwriting and marketing aiSure.
Jurisdiction: German regulated entity. Global reach via Mosaic Insurance partnership.
Capacity: Up to EUR/USD/CAD 15 million in initial capacity via Mosaic Insurance, which underwrites and markets Munich Re aiSure.
Target segment: Mid-market to enterprise. AI developers, vendors, and deployers.
Underwriting posture: Performance-based. Munich Re publishes that aiSure settles on measurable performance data rather than traditional loss adjustment, and that it requires Munich Re technical due diligence. Munich Re does not use that word for it anywhere it publishes, and neither does this desk.
Coverage triggers: Algorithmic bias producing decisions that discriminate on protected characteristics; privacy failures exposing personal or confidential data; intellectual property infringement from AI outputs; performance shortfall where the system underperforms against agreed accuracy, hallucination rate, or uptime thresholds.
Exclusions: Losses below the agreed performance trigger threshold; losses in unmeasured categories not specified in the policy; regulatory penalties under most European regimes. Basis risk, meaning actual loss exceeding the measured shortfall, is structural rather than an exclusion.
Geographic reach: Global, including European operators.
Certification references: Not published by Munich Re at the level of a named standard. Governance documentation aligned to ISO/IEC 42001 and the NIST AI Risk Management Framework is what documentation-intensive underwriting of this kind generally asks for, and that is this desk's reading rather than a Munich Re statement.
Minimum premium: Not publicly disclosed. Underwriting is documentation-intensive; expect multi-week timelines for initial placement.
Sources: munichre.com and mosaicinsurance.com (checked 17 August 2026).

HSB AI Liability Insurance

Product name: AI Liability Insurance for Small Businesses
Parent / capital: Hartford Steam Boiler Inspection and Insurance Company (HSB), a Munich Re subsidiary. HSB is a specialty insurer focused on equipment breakdown, technology, and specialty risk.
Launch: 18 March 2026. Announced via Business Wire from Hartford, Connecticut.
Jurisdiction: US. Subject to regulatory approval in individual states. Distribution through insurance carrier partners that incorporate HSB coverage into their commercial products; HSB does not sell directly to businesses.
Capacity: Structured within commercial general liability frameworks. No standalone per-occurrence limit publicly disclosed; coverage is an add-on to carrier partner policies.
Target segment: Small and medium-sized businesses. HSB's own survey found 74% of SMBs using AI programmes as of early 2026.
Underwriting posture: Indemnity. Extension to existing commercial policies through carrier partner distribution.
Coverage triggers: (1) Bodily injury liability: lawsuits alleging a person was injured due to the insured's use of AI, including AI-controlled physical systems such as HVAC equipment. (2) Property damage liability: lawsuits claiming property was damaged due to AI use, including AI-generated instructions causing physical damage. (3) Advertising injury: claims arising from AI-generated advertising, marketing copy, blog content, and social media posts, including defamation, copyright infringement in AI-generated content, and invasion of privacy.
Exclusions: Not publicly detailed beyond standard commercial lines exclusions. Coverage is specifically framed as addressing losses that "some General Liability policies exclude" due to AI use.
Geographic reach: United States, subject to state regulatory approval.
Certification references: None publicly specified. Product is designed for accessibility rather than governance-intensive underwriting.
Sources: Business Wire press release 18 March 2026; insurance-canada.ca; InsuranceNewsNet; completeaitraining.com reporting.

Armilla AI Liability Policy

Product name: AI Liability Insurance (standalone); Vanguard AI (combined framework with Chaucer)
Parent / capital: Armilla Insurance Services, a Coverholder at Lloyd's. First Lloyd's coverholder dedicated exclusively to AI liability insurance. Cover underwritten by certain underwriters at Lloyd's, including Chaucer.
Launch: Armilla was the first Lloyd's coverholder dedicated exclusively to AI liability, in 2024. Vanguard AI, the coordinated cyber, technology and AI liability structure built with Chaucer, launched 10 February 2026. A 30 April 2025 product launch date and a January 2026 capacity announcement appeared in an earlier version of this entry sourced to trade press; neither could be confirmed at armilla.ai and both have been removed. The USD 25 million figure is a policy limit, not a funding round.
Jurisdiction: Surplus lines insurance. Available through licensed surplus lines brokers. Geographic availability varies by jurisdiction. Lloyd's regulatory framework applies to the London underwriting component.
Capacity: Up to USD 25 million per organisation.
Target segment: Enterprises deploying AI solutions, particularly those with internally developed or customised models. Mid-market to large enterprise.
Underwriting posture: Indemnity, monoline specialty. Affirmative trigger around underperformance of AI applications.
Coverage triggers: Hallucinations and model drift; inaccurate AI outputs causing third-party loss; data leakage from AI systems; AI regulatory violations; non-breach privacy incidents; AI model error liability; harmful AI agent outputs; AI-driven property damage; defence costs under AI regulations including the EU AI Act and US state AI laws (Colorado AI Act).
Common exclusions: Not publicly specified. Standard surplus lines conditions apply. Chaucer has noted selectivity: "We will be selective. We won't cover AI systems that are excessively prone to breakdown."
Geographic reach: United States (surplus lines) and select international markets. Available to European buyers through wholesale broker intermediaries, though not a European-domiciled product.
Certification references: Collaboration with Trustible announced 8 October 2025. No named certification standard is published as an underwriting reference.
Sources: Armilla press release April 2025; prnewswire.com; ffnews.com January 2026; Reinsurance News; CDO Magazine; Lloyd's Lab alumni profile.

AIUC (Artificial Intelligence Underwriting Company)

Product name: AIUC-1 (certification standard; AIUC is not an insurer)
Parent / capital: Independent, US-incorporated. USD 15 million seed round led by Nat Friedman at NFDG.
Launch: Emerged from stealth July 2025. AIUC-1 standard published mid-2025. Certificates run twelve months with at least quarterly testing.
Jurisdiction: United States. San Francisco, California.
Capacity: Not applicable; AIUC publishes no insurance capacity.
Target segment: Enterprise buyers evaluating AI agent deployments. Security, legal, and procurement functions at organisations deploying third-party AI agents.
Underwriting posture: Standards-led. AIUC-1 is built to connect certification to insurance; the first AIUC-1-backed policy, written for ElevenLabs, was placed through Lloyd's of London. The standard is positioned as "SOC 2 for AI agents."
Coverage triggers: Not applicable to the standard. The terms of the ElevenLabs policy are not disclosed.
Common exclusions: Not publicly specified.
Geographic reach: United States. No explicit international expansion announced.
Certification references: AIUC-1 (proprietary). Operationalises ISO/IEC 42001, the NIST AI RMF, MITRE ATLAS and the OWASP Top 10 for LLMs. Audit services provided through accredited partners including Schellman.
Sources: Fortune July 2025; prnewswire.com; Reinsurance News; Emergence Capital blog; calpcc.com on AIUC-1; aiuc.com.

Testudo GenAI Liability Insurance

Product name: GenAI Liability Insurance
Parent / capital: Testudo, independent US startup. Backed by Lloyd's Lab and tier-one investors. Lloyd's capacity (Apollo and other syndicates) provides the underwriting paper.
Launch: 21 January 2026. Capacity expanded to USD 9.25 million per insured on 26 February 2026, when Atrium and QBE joined Apollo.
Jurisdiction: United States. Claims-made policy form.
Capacity: Up to USD 9.25 million per insured since 26 February 2026.
Target segment: US enterprises deploying generative AI systems. Middle to large enterprise focus.
Underwriting posture: Indemnity, monoline specialty. Proprietary methodology combining real-time litigation data and external risk signals. No invasive technical audit required at application; faster quote turnaround is a stated competitive advantage.
Coverage triggers: Hallucinations; intellectual property infringement from AI outputs; physical property damage caused by AI; unauthorised data disclosures from AI systems; comprehensive liability for generative AI-caused third-party damages. Specifically designed to respond where Commercial General Liability exclusions effective January 2026 left gaps.
Common exclusions: Product is designed to address the gap created by ISO CG 40 47 and CG 40 48 exclusions. Specific policy exclusions not publicly disclosed.
Geographic reach: United States.
Certification references: Proprietary active litigation monitoring technology rather than certification-based underwriting. Testudo monitors US litigation data sets to understand where GenAI liability risk is concentrating.
Sources: testudo.co launch announcement January 2026; Fintech Global March 2026; S&P Global Market Intelligence February 2026; Reinsurance News; Insurance Times.

Counterpart Affirmative AI Coverage

Product name: Affirmative AI Coverage (extension across professional liability lines)
Parent / capital: Counterpart, US insurtech. Management liability and professional liability specialist.
Launch: Announced 21 November 2025 via Business Wire. Available from launch date across Miscellaneous Professional Liability, Allied Health, and Technology E&O product lines.
Jurisdiction: United States.
Capacity: Professional liability limits; typical range not separately disclosed for AI extensions. Standard professional liability limits for the applicable line apply.
Target segment: Mid-market professional services firms, consultants, financial advisors, insurance agents, healthcare-adjacent technology, and technology companies with E&O exposure.
Underwriting posture: Indemnity, extension model. AI coverage added as an insuring agreement within existing professional liability policy structures. Technology E&O insuring agreement added as part of the expansion.
Coverage triggers: Hallucinated reports causing client loss; AI-misclassified risk exposures leading to client disputes; AI-generated hiring recommendations causing discrimination claims; AI decision errors in professional advice contexts; AI-assisted assessments producing flawed projections. Language specifically addresses AI-generated mistakes, decision errors, and biased outputs.
Common exclusions: Standard professional liability exclusions apply. Coverage is framed as affirmative for AI-specific triggers within the existing policy structure.
Geographic reach: United States.
Certification references: Not publicly specified. Underwriting through Counterpart's standard professional liability process.
Sources: Business Wire 21 November 2025; Yahoo Finance; InsuranceNewsNet; Hunton Andrews Kurth Affirmative AI coverages analysis; National Law Review.

Corgi Insurance

Product name: AI Liability Insurance (within full startup insurance platform)
Parent / capital: Corgi Insurance, San Francisco. AI-native insurance platform for startups. Corgi publishes that Corgi Insurance Services, Inc. is a licensed insurance producer, CA licence number 6012791, acting as programme administrator rather than as the insurer. Read that carefully before treating Corgi as a carrier: it changes who carries the risk on your policy. Corgi announces a USD 108 million fundraise on its own site. Investor names, revenue figures and a regulatory approval date appeared in an earlier version of this entry sourced to trade press; none is published by Corgi and all have been removed.
Launch: Founded 2024.
Jurisdiction: United States. Licensed insurance carrier.
Capacity: Not separately disclosed for AI liability line. Part of integrated startup insurance platform.
Target segment: Venture-backed and high-growth technology companies. AI liability sits within a broader platform covering D&O, E&O, CGL, cyber, HNOA, fiduciary, and representations and warranties coverage.
Underwriting posture: Indemnity, through a programme administrator model. Corgi designs and manages the programme end to end using its own underwriting, policy management and claims infrastructure.
Coverage triggers: Corgi publishes that Tech and AI Liability covers claims alleging that the insured's technology products or services failed to perform as intended and caused a client financial harm, naming hallucinated, inaccurate or allegedly biased model outputs. It also publishes that the cover may sit inside Technology E&O and is subject to terms, exclusions, limits and endorsements.
Common exclusions: Standard technology company insurance exclusions. Full policy terms available through the Corgi platform.
Geographic reach: United States.
Certification references: Underwriting through Corgi's proprietary AI platform. Certification references for AI liability line not publicly specified.
Sources: corgi.insure/ai (checked 17 August 2026).

Vouch AI Insurance

Product name: AI Insurance (tech E&O with AI extensions)
Parent / capital: Vouch Insurance, San Francisco. Hiscox announced acquisition of Corix, the underwriting division of Vouch, bringing Hiscox capacity behind Vouch products.
Launch: AI-specific product page and coverage available through 2024-2026 product evolution. Hiscox Corix underwriting partnership active as of 2026.
Jurisdiction: United States. Coverage availability varies by state.
Capacity: Not publicly disclosed. Structured within tech E&O policy limits.
Target segment: AI startups. Platform serves 800+ AI startup clients as of 2026.
Underwriting posture: Indemnity, extension model. AI coverage integrated within tech E&O and related startup insurance product lines.
Coverage triggers: IP infringement from AI training data or model outputs; bias and discrimination claims from algorithmic decisions; regulatory investigation defence costs; errors and omissions specific to AI model performance; generative output copyright and media liability; enterprise indemnity and contractual liability; data misuse disputes; D&O coverage included in startup package.
Common exclusions: Not specifically disclosed beyond standard tech E&O conditions. Same-day quoting available through platform.
Geographic reach: United States. Not all products available in all states.
Certification references: Not specified. Underwriting through Corix (Hiscox) process.
Sources: vouch.us/technology/ai; Vouch blog on E&O vs. AI insurance; Fintech Global reporting on Cara USD 8 million seed (adjacent market context).

Section 4. The Retreating Generalists

While specialist carriers are building affirmative products, the generalist market moved in the opposite direction beginning in late 2025 and accelerating through Q1 2026. Understanding the exclusion landscape is as important as understanding available coverage, because a buyer without explicit affirmative AI coverage may discover that their existing policies exclude AI losses at renewal.

The ISO Verisk Exclusion Endorsements

Updated 17 August 2026.

CG 40 47, CG 40 48 and CG 35 08, their edition dates and their January 2026 effective date are reported by trade press and law-firm analysis of the ISO multistate general liability filing. We have not confirmed them against the filed forms at ISO, at Verisk, or through a state insurance filing portal; the most recent attempt was on 17 August 2026. ISO forms are a subscription product, so the absence of a public record is not evidence that a form does not exist. It does mean these numbers are reported rather than established. The numbers are kept here because they are how a reader finds the endorsement on their own schedule. Ask your broker to confirm the exact form number and edition date on that schedule before relying on any of the analysis below. A search engine will return a confident answer confirming these numbers; that answer is assembled in part from pages published by this network, this one included.

ISO, the standard-forms business now part of Verisk, is reported to have introduced two optional endorsement forms available to carriers from 1 January 2026:

  • CG 40 47: Exclusion of Generative Artificial Intelligence. Broad exclusion applicable to both Coverage A (bodily injury and property damage) and Coverage B (personal and advertising injury) under the ISO Commercial General Liability Coverage Part. Excludes losses arising out of generative artificial intelligence.
  • CG 40 48: Exclusion of Generative Artificial Intelligence (Coverage B Only). Applies the exclusion only to personal and advertising injury claims under Coverage B, retaining Coverage A for bodily injury and property damage claims with an AI nexus. A narrower option for carriers that want to limit advertising injury exposure without removing bodily injury coverage.

The form CG 40 48 is available as a PDF from ALM and has been reviewed by insurance law publications. Both forms are reported to have been the subject of a July 2025 ISO multistate filing covering emerging risks in general liability. Carrier interest ahead of the reported January 2026 effective date is described as strong in the same trade coverage. Neither the interest nor the date is confirmed at Verisk.

Carriers Filing Exclusions

W.R. Berkley proposed an absolute exclusion that would bar claims tied to "any actual or alleged use" of AI, even where AI forms only a minor component of a product or workflow. The proposed Berkley language specifically names individual AI tools by name, including ChatGPT, Bard, Midjourney, and DALL-E, creating product-specific exclusion language that goes further than the ISO standard forms.

AIG filed AI exclusion language with state regulators covering management liability and other professional lines. AIG told regulators it had no plans to implement the exclusions immediately but wanted the language approved and available as claims frequency increases. AIG, Great American and Philadelphia Indemnity are among the carriers reported to have submitted AI-specific exclusion filings with US state regulators.

Great American Insurance has also sought regulatory clearance for AI-specific exclusions in management liability policies. Hamilton Insurance Group is reported in trade and law firm analysis to have filed AI exclusion language, but the accounts disagree on whether it is a management liability exclusion or a sublimit inside errors and omissions or cyber cover, and Hamilton publishes nothing about AI underwriting on its own site.

Chubb has taken a more nuanced position, agreeing to cover certain AI-related incidents while introducing an exclusion for events capable of affecting large numbers of insureds simultaneously (systemic AI events). Chubb's approach distinguishes isolated incident exposure from correlated catastrophic AI exposure.

London market specialists active in cyber and professional liability are reported to be introducing AI sublimits rather than full exclusions. UK firms renewing cyber cover in 2026 are encountering policy schedules that itemise AI sublimits explicitly. The percentage caps that circulated in the trade press are not published by any carrier, checked 17 August 2026, and none is stated here. What is worth acting on is the structure rather than the number: if your schedule itemises an AI sublimit, the figure on your own schedule is the only one that governs your claim.

The common rationale across all retreating generalists is identical: AI risk is structurally different from the risks these policies were priced to cover, the loss distribution is unknown, and the potential for correlated large losses is not reflected in existing premium rates. The carriers filing exclusions are not making a judgment that AI losses will not occur. They are making a judgment that they cannot currently price those losses.

Section 5. Lloyd's and the Wholesale/Coverholder Layer

Lloyd's of London is the market infrastructure through which AI liability capacity reaches buyers who cannot access it through US licensed carriers or London company market insurers. The Lloyd's market operates through syndicates that underwrite risk on subscription, through managing agents that run those syndicates, and through coverholders that are authorised to bind coverage on behalf of syndicates within defined parameters.

For AI liability, Lloyd's is performing its traditional function: providing capacity for a new and poorly modelled risk class where the admitted market has not yet established a position. The Lloyd's Lab accelerator programme has supported multiple AI insurance startups including Testudo, providing both capital access and underwriting expertise.

Chaucer: The Most Active AI Syndicate

Chaucer Group is the Lloyd's syndicate most publicly active in AI liability underwriting as of April 2026. Chaucer underwriters the Armilla AI Liability Policy and co-developed the Vanguard AI combined framework with Armilla, which integrates cyber, technology, and AI liability coverage in a single structured policy. Tom Graham, Chaucer's representative on AI underwriting, has described the approach publicly: affirmative AI coverage is available but selective. Chaucer will not cover AI systems that are structurally prone to frequent failures.

Apollo and Other Syndicates: Testudo Capacity

Apollo (an integrated global alternative asset manager, separate from the Lloyd's syndicate structure but active in providing Lloyd's paper) and other Lloyd's syndicates provide the underwriting capacity behind Testudo's GenAI Liability Insurance programme. The Lloyd's Lab backing of Testudo represents the Lloyd's market's institutional endorsement of the programme as a credible approach to the GenAI liability class.

Access Routes for European Buyers

European buyers accessing Lloyd's AI capacity typically do so through one of three routes. First, through a surplus lines or non-admitted basis where the European policyholder accesses a US-licensed surplus lines carrier that places the risk with Lloyd's syndicates. Second, through a European wholesale broker that has established relationships with Chaucer, Apollo, or other Lloyd's underwriters and can place European risk directly. Third, through the Lloyd's Brussels platform, which has an EU passporting arrangement for certain classes. None of these routes provides coverage under EU-domiciled policy language, which remains the gap.

Section 6. Capacity by Segment

The following tables summarise carriers by target segment and indicative capacity ranges. Figures are market observations drawn from public announcements and trade reporting. Available terms, actual limits, and pricing will vary based on underwriting review of specific risks. This is not a guarantee of available coverage.

SME Segment (revenues below USD 10 million)

Carrier Product Capacity (indicative) Distribution
HSB (Munich Re subsidiary) AI Liability Insurance CGL-embedded; limits vary by carrier partner policy Via insurance carrier partners; not direct
Vouch (Hiscox/Corix backing) AI Insurance (tech E&O extension) Tech E&O limits; not separately disclosed Direct platform; same-day quoting
Corgi (full-stack carrier) AI Liability within startup package Not separately disclosed Direct platform

Mid-market segment (this desk's band: revenues roughly USD 10 million to USD 500 million)

Carrier Product Capacity (indicative) Distribution
Testudo GenAI Liability Insurance Up to USD 9.25 million per insured Broker; Lloyd's-backed
Counterpart Affirmative AI Coverage Professional liability limits (typically USD 5M to 15M) Broker; US admitted market
Munich Re aiSure aiSure (AI performance cover) Up to EUR/USD/CAD 15 million in initial capacity via Mosaic Broker with Mosaic relationship; global

Enterprise Segment (revenues above USD 500 million, or AI at regulated scale)

Carrier Product Capacity (indicative) Distribution
Armilla (Chaucer / Lloyd's) AI Liability Policy; Vanguard AI Up to USD 25 million per organisation Surplus lines broker; Lloyd's wholesale
Munich Re aiSure aiSure (AI performance cover) Up to EUR/USD/CAD 15 million initial capacity Mosaic Insurance
AIUC AIUC-1 certification standard (not a carrier) Not publicly disclosed Standard; the ElevenLabs policy was placed through Lloyd's
Lloyd's bespoke (layered) Brokered programme across syndicates Above USD 50 million is achievable layered, on this desk's reading of published single-carrier limits. No syndicate publishes an AI aggregate. Wholesale Lloyd's brokers; requires strong documentation

Section 7. Underwriting Criteria Convergence

Despite the diversity of products in this market, the underwriting criteria that AI liability carriers reference are converging around five core factors. A buyer that understands these five factors can prepare a submission that works across multiple carriers simultaneously, rather than customising documentation for each carrier separately.

Factor 1: Certification Posture

A carrier with a documentation-intensive underwriting process is likely to ask some version of the same question: has this AI system been assessed against a recognised framework? The frameworks most consistently referenced are:

  • NIST AI Risk Management Framework (AI RMF 1.0), released 26 January 2023, NIST AI 100-1 (nist.gov, checked 17 August 2026). A voluntary framework, and the one most consistently named in AI underwriting submissions as baseline evidence of structured risk governance. The Generative AI Profile, NIST AI 600-1, released 26 July 2024, supplements it for LLM deployments.
  • ISO/IEC 42001:2023, published December 2023. AI management system requirements, and the AI counterpart to ISO 27001 for information security. Its technical documentation requirements overlap substantially with EU AI Act Articles 9 and 11, which is why one evidence set can serve both.
  • AIUC-1, published by the Artificial Intelligence Underwriting Company, which came out of stealth in July 2025. A standard of 51 requirements and 130 controls across six pillars: data and privacy, security, safety, reliability, accountability and societal risks. Specifically designed for AI agents. Audits conducted through accredited partners including Schellman. Represents the most AI-agent-specific framework currently available.
  • SOC 2 Type II. An information security attestation, often read as a proxy for organisational security and governance maturity.
  • Agent Certified (agentcertified.eu). Seven-dimension EU-aligned certification framework published by Future Proof Intelligence, which also publishes this desk. It is designed for the EU regulatory context, including EU AI Act and Product Liability Directive compliance mapping. No carrier has adopted it as a named underwriting reference, and we will not imply otherwise on a page whose whole purpose is to be relied on.

Factor 2: Autonomy Envelope

How much can the AI system do without a human in the loop? Carriers universally apply higher loadings as autonomy increases. An AI tool that generates a draft for human review is underwritten differently from an AI agent that executes transactions autonomously. The scope of the autonomy envelope, the controls that constrain it, and the rollback mechanisms available all affect terms.

Factor 3: Deployment Scope

How many people does the AI system interact with, and in what capacity? A system deployed internally to assist 50 analysts is underwritten differently from a system deployed to interact with millions of consumers. Carriers assess the maximum affected population in a failure scenario as a component of aggregate loss modelling.

Factor 4: Sector Sensitivity

AI systems deployed in sectors listed in EU AI Act Annex III (healthcare, education, employment, essential services, law enforcement, border control, administration of justice, and critical infrastructure) attract higher loadings across all carriers. US carriers reference equivalent sector sensitivities even without formal Annex III language, given the concentration of litigation in healthcare, financial services, and employment contexts.

Factor 5: Claims History

Given that the market is new, few buyers have AI-specific claims history. Carriers use analogous history from technology errors and omissions, cyber liability, and product liability as proxies. First-time AI liability buyers may meet a premium loading for the absence of positive claims experience. As the market matures, a clean claims record on AI-specific policies will become a meaningful underwriting signal.

These five factors are why governance documentation, third-party validation, and autonomy controls belong in a submission: they give every underwriter approached the same evidence to read. Whether that reduces a premium loading is each carrier's decision, and none has published a figure for it.

Section 8. Geographic Access

The following matrix maps carrier access by geography. "Available" means the carrier has publicly indicated it writes the relevant geography or the coverage is structurally accessible. "Via wholesale" means access requires a Lloyd's wholesale or surplus lines broker intermediary. "Pending" means the carrier has indicated interest but no confirmed market position.

Carrier / Product US EU UK APAC
Munich Re aiSure Available Available Available Available
HSB AI Liability Insurance Available Not available Not available Not available
Armilla / Chaucer (Lloyd's) Available Via wholesale Via wholesale Via wholesale
AIUC Available Not confirmed Not confirmed Not confirmed
Testudo Available Not confirmed Not confirmed Not confirmed
Counterpart Available Not available Not available Not available
Corgi Available Not available Not available Not available
Vouch (Hiscox/Corix) Available Not confirmed Not confirmed Not confirmed
Lloyd's bespoke programme Available Via wholesale Available Via wholesale

The geographic access picture underscores the structural gap for European buyers. With the EU AI Act enforcement date of 2 August 2026 approaching and Product Liability Directive transposition required by 9 December 2026, European operators have regulatory exposure that existing US-domiciled products do not specifically address. The coverage is accessible via wholesale routing, but the policy language, underwriting criteria, and claims frameworks are not designed around the EU regulatory environment.

Section 9. What the Market Will Look Like in 2027

The structural forces shaping the 2026 market will produce a materially different market by 2027. Four developments are likely.

Consolidation among specialty carriers. The current cohort of specialists (the Lloyd's coverholders Armilla and Testudo, and Corgi) is competing for the same enterprise buyer segment with limited actuarial data. As the first generation of policies runs through its first claims cycle, the carriers with the strongest underwriting frameworks and the cleanest loss experience will attract the most capacity. Carriers with weaker governance and higher loss ratios will find reinsurance support tightening. By 2027, the five or six specialty carriers that have survived the first claims cycle will have meaningful advantages in pricing, capacity, and broker relationships.

Lloyd's syndicate commitment deepening. Chaucer's public commitment to AI underwriting and Lloyd's Lab's institutional support for Testudo represent the leading edge of a deeper Lloyd's market position. As reinsurance capacity develops behind the specialist primary carriers, Lloyd's syndicates will have greater confidence to commit capacity. By 2027, expect multiple named syndicates with published AI liability underwriting appetites, rather than the current position where most Lloyd's capacity flows through a single coverholder relationship.

Traditional carriers re-entering on narrow terms. The carriers currently filing exclusions are doing so because they cannot model the risk. As the first generation of AI liability data emerges from specialty carrier experience, traditional carriers will begin re-entering the market on carefully defined terms: specific sectors, defined use cases, sub-limits within existing policy structures. The bifurcation will partially reverse, but the terms on which traditional carriers re-enter will be narrower than the silent coverage that the ISO exclusions removed.

A European-native product. The absence of a European-domiciled primary AI insurer is the most significant market gap in 2026. By 2027, at least one European carrier or Lloyd's Brussels vehicle is likely to have launched a product with EU-specific policy language, incorporating EU AI Act Article 26 operator obligations and Directive 2024/2853 product liability triggers as defined coverage triggers rather than general professional liability language. EIOPA's Opinion on AI governance and risk management of 6 August 2025, and its survey on generative AI of 2 February 2026, are the supervisory architecture that will enable this. This publication will track it.

Section 10. Frequently Asked Questions

The map above is a snapshot of structure. For the maintained version, re-verified at each carrier's own domain with the date on every line, see the carrier matrix and its machine-readable companion. For the buyer-facing walkthrough of the same market, the complete European AI agent insurance market guide.

Questions

Who offers AI liability insurance in 2026?

As of April 2026, the providers writing meaningful AI liability limits include Munich Re (aiSure, AI performance cover), Armilla (Lloyd's coverholder, up to USD 25 million), Testudo (Lloyd's coverholder, USD 9.25 million per insured since 26 February 2026), the AIUC-1 standard from the Artificial Intelligence Underwriting Company (a standards body, not a carrier), Counterpart (affirmative AI coverage across professional liability), HSB (Munich Re subsidiary, SME cover introduced March 2026 and added to partner carriers' business policies), Corgi (full-stack AI-native carrier for tech startups), and Vouch (tech E&O extensions, distributed via Hiscox Corix). Additional Lloyd's capacity is available through wholesale broker intermediaries. The market is bifurcated: specialist carriers are building products while mainstream generalists are adding AI exclusions.

Which carriers have AI exclusions in 2026?

AIG, W.R. Berkley and Great American are reported to have filed AI exclusion language with US state regulators; the filings themselves are not public and none of the three publishes the wording. Hamilton Insurance Group is reported to have done the same, though the line of business is not confirmed at source. London market specialists are reported to be introducing AI sublimits on cyber policies. No carrier publishes a percentage cap and this desk states none. Chubb is reported to have excluded widespread or systemic AI events. The ISO Verisk endorsement forms CG 40 47 (broad, Coverage A and B) and CG 40 48 (Coverage B only) became available to all carriers from January 2026, and many have implemented them at renewal.

Can an SME buy AI agent insurance in 2026?

Yes. HSB launched a purpose-built AI Liability Insurance product for small and medium-sized businesses in March 2026, distributed through insurance carrier partners. It covers bodily injury, property damage, and advertising injury arising from AI use. Vouch and Corgi also serve smaller technology companies through tech E&O platforms. The primary gap for SMEs is access to higher limits: specialist products above USD 5 million are likely to ask for governance documentation that smaller organisations may not have prepared.

What is HSB AI Liability Insurance?

HSB AI Liability Insurance is a commercial product launched by Hartford Steam Boiler (a Munich Re subsidiary) on 18 March 2026 for small and medium-sized businesses. It covers bodily injury liability from AI use (including AI-controlled physical systems), property damage from AI-generated instructions, and advertising injury from AI-generated marketing and social media content. Distributed through carrier partners rather than directly. Pending regulatory approval by state. Designed to address losses that standard CGL policies may exclude due to AI use.

How does Munich Re aiSure work?

Munich Re aiSure is an AI performance insurance product distributed via Mosaic Insurance. It settles when a pre-agreed performance trigger is breached: model accuracy below an agreed threshold, hallucination rate above a defined level, or uptime below a minimum. Coverage includes algorithmic bias, privacy failures, intellectual property infringement, and performance shortfalls. Capacity via Mosaic is up to EUR/USD/CAD 15 million in initial capacity. Mosaic Insurance underwrites and markets aiSure. Global reach, including European operators.

What does Armilla cover?

Armilla is a Coverholder at Lloyd's and the first dedicated exclusively to AI liability. Its policy covers hallucinations, model drift, inaccurate outputs, data leakage, AI regulatory violations, non-breach privacy incidents, AI agent failures, AI-driven property damage, and defence costs under AI regulations including the EU AI Act. Coverage limits reach up to USD 25 million per organisation. Chaucer is the lead Lloyd's syndicate. Available on a surplus lines basis through licensed surplus lines brokers.

Is AIUC insurance or a certification standard?

The Artificial Intelligence Underwriting Company (AIUC) is an underwriting company and standards body. It launched from stealth in July 2025 with a USD 15 million seed round led by Nat Friedman at NFDG. AIUC-1 is its security, safety and reliability standard for AI agents: 51 requirements and 130 controls across six pillars, namely data and privacy, security, safety, reliability, accountability, and societal risks. It operationalises ISO/IEC 42001, the NIST AI RMF, MITRE ATLAS and the OWASP Top 10 for LLMs. The standard is built to connect certification to insurance: the first AIUC-1-backed AI agent policy, written for ElevenLabs in February 2026, was placed through Lloyd's of London. How a certificate changes the terms of a policy is not published.

Do Lloyd's syndicates write AI liability insurance?

Yes. Chaucer is the most publicly active Lloyd's syndicate, underwriting Armilla's standalone AI Liability Policy and co-developing the Vanguard AI combined framework. Apollo and other syndicates provide capacity behind Testudo's programme. Lloyd's Lab has institutionally supported AI insurance startups including Testudo. European buyers access Lloyd's AI capacity through surplus lines brokers, wholesale Lloyd's intermediaries, or the Lloyd's Brussels EU passporting vehicle.

How much AI liability capacity is available per risk in 2026?

SME buyers: CGL-embedded via HSB carrier partners. Mid-market: USD 9.25 million per insured (Testudo), Counterpart writes AI cover inside its professional liability limits and does not disclose a separate AI range; up to EUR/USD/CAD 15 million (Munich Re aiSure via Mosaic). Enterprise: up to USD 25 million (Armilla), bespoke programmes above USD 50 million are plausible through layered Lloyd's placements, on this desk's arithmetic from published single-carrier limits. Per-risk capacity continues to expand as the market develops underwriting experience.

What certifications do AI insurers require?

No single certification is universally required, and no carrier publishes a list of the frameworks it recognises. The instruments most often named in AI governance submissions are the NIST AI Risk Management Framework (AI RMF 1.0) and ISO/IEC 42001:2023 (AI management systems). AIUC-1, published by the Artificial Intelligence Underwriting Company, is the most AI-agent-specific standard and the one built to connect certification to insurance: the first AIUC-1-backed policy, written for ElevenLabs, was placed through Lloyd's of London. SOC 2 Type II is often read as a proxy for security and governance maturity. Agent Certified (agentcertified.eu), a seven-dimension EU-aligned framework published by Future Proof Intelligence, is written so an underwriter can read it, but no carrier has adopted it as a named underwriting reference and no insurer has committed to a change in terms for holding a result.

Sources

  1. Hartford Steam Boiler Inspection and Insurance Company (HSB). "HSB Introduces AI Liability Insurance for Small Businesses." Press release. Business Wire, 18 March 2026. https://www.businesswire.com/news/home/20260318144322/en/HSB-Introduces-AI-Liability-Insurance-for-Small-Businesses
  2. Hartford Steam Boiler Inspection and Insurance Company (HSB). Munich Re Group. "HSB AI Liability Insurance for Small Businesses." munichre.com/hsb, 18 March 2026. https://www.munichre.com/hsb/en/press-and-publications/press-releases/2026/2026-03-18-introducing-ai-liability-insurance-for-small-businesses.html
  3. Armilla Insurance Services. Affirmative AI Liability Insurance, underwritten by certain underwriters at Lloyd's; partners named at source are Chaucer Group, AXIS Capital, Convex, Greenlight Re and Swiss Re. armilla.ai (checked 17 August 2026). Trade announcement retained for the record: PR Newswire, https://www.prnewswire.com/news-releases/armilla-launches-affirmative-ai-liability-insurance-with-lloyds-underwriter-chaucer-302442586.html
  4. Armilla. Standalone AI Liability Policy, limits of up to USD 25 million per organisation. https://www.armilla.ai/ai-insurance
  5. Armilla AI. Lloyd's Lab alumni profile. Lloyd's of London. https://www.lloyds.com/insights/lloyds-lab/programmes-and-initiatives/lloyds-lab-accelerator/alumni/armilla-ai
  6. The Artificial Intelligence Underwriting Company (AIUC). "The Artificial Intelligence Underwriting Company launches with $15M to help enterprises deploy AI with confidence." PR Newswire, July/August 2025. https://www.prnewswire.com/news-releases/the-artificial-intelligence-underwriting-company-launches-with-15m-to-help-enterprises-deploy-ai-with-confidence-302512447.html
  7. Fortune. "AIUC, a startup creating insurance for AI agents, emerges from stealth with $15 million seed." 23 July 2025. https://fortune.com/2025/07/23/ai-agent-insurance-startup-aiuc-stealth-15-million-seed-nat-friedman/
  8. Emergence Capital. "AIUC: Establishing Trust and Risk Standards for Enterprise AI." https://www.emcap.com/thoughts/aiuc-establishing-trust-risk-standards-for-enterprise-ai
  9. Testudo. "Insurance for GenAI Liability Risks Now Available at Testudo." Launch announcement, 21 January 2026. https://www.testudo.co/insights/testudo-launches-new-insurance-coverage-for-liability-risks-created-by-generative-ai-systems
  10. Testudo. "Lloyd's syndicates commit more capacity to generative AI liability insurance." 26 February 2026. https://www.testudo.co/insights/lloyd-s-syndicates-commit-more-capacity-to-generative-ai-liability-insurance
  11. S&P Global Market Intelligence. "As insurers retreat from AI risk, one startup plans to fill the gap." February 2026. https://www.spglobal.com/market-intelligence/en/news-insights/articles/2026/2/as-insurers-retreat-from-ai-risk-one-startup-plans-to-fill-the-gap-97375264
  12. Counterpart. "Leading Insurtech, Counterpart, Addresses Critical Coverage Gap With Affirmative AI Coverage." Business Wire, 21 November 2025. https://www.businesswire.com/news/home/20251121123510/en/Leading-Insurtech-Counterpart-Addresses-Critical-Coverage-Gap-With-Affirmative-AI-Coverage
  13. Corgi Insurance. Tech and AI Liability product page and company description. corgi.insure/ai (checked 17 August 2026).
  14. Verisk (Insurance Services Office). CG 40 47 and CG 40 48 endorsement forms, effective 1 January 2026. Described in: Verisk Core. "Emerging Risks in ISO General Liability Multistate Filing." July 2025. https://core.verisk.com/Insights/Emerging-Issues/Articles/2025/July/Week-4/Emerging-Risks-in-ISO-General-Liability-Multistate-Filing
  15. CG 40 48 form text. "Exclusion of Generative Artificial Intelligence (Coverage B Only)." ALM. https://assets.alm.com/63/68/46ed4bf34a0e807c9695e15c9e19/cg-40-48-01-26-exclusion-generative-artificial-intelligence-coverage-b-only.pdf
  16. Reinsurance News. "Artificial Intelligence Underwriting Company launches with $15m seed round." July 2025. https://www.reinsurancene.ws/artificial-intelligence-underwriting-company-launches-with-15m-seed-round/
  17. Reinsurance News. "Armilla reveals purpose-built AI liability insurance amid rising legal and regulatory pressures." https://www.reinsurancene.ws/armilla-reveals-purpose-built-ai-liability-insurance-amid-rising-legal-and-regulatory-pressures/
  18. Insurance Times. "Startup to launch new AI underwriting platform as it looks to tackle 'impossible market'." https://www.insurancetimes.co.uk/news/startup-to-launch-new-ai-underwriting-platform-as-it-looks-to-tackle-impossible-market/1455409.article
  19. Moffatt v. Air Canada. Civil Resolution Tribunal, British Columbia, Case No. 2024-00451. Decision issued 14 February 2024. Confirmed company liability for negligent misrepresentation by AI chatbot. McCarthy Tétrault analysis: https://www.mccarthy.ca/en/insights/blogs/techlex/moffatt-v-air-canada-misrepresentation-ai-chatbot
  20. Mata v. Avianca, Inc. No. 1:2022-cv-01461, S.D.N.Y. 2023. Sanctions imposed following submission of AI-hallucinated case citations. Document 54, Justia. https://law.justia.com/cases/federal/district-courts/new-york/nysdce/1:2022cv01461/575368/54/
  21. European Insurance and Occupational Pensions Authority (EIOPA). Survey on GenAI use in the European insurance sector. February 2026. Frankfurt.
  22. Regulation (EU) 2024/1689 (the EU AI Act). Articles 9, 11, 26, 72, and 99. Official Journal of the European Union, 12 July 2024.
  23. Directive (EU) 2024/2853 (revised Product Liability Directive). OJ L, 18 November 2024. Transposition deadline 9 December 2026.
  24. International Organization for Standardization. ISO/IEC 42001:2023. Artificial Intelligence Management Systems. Geneva, December 2023. https://www.iso.org/standard/81230.html
  25. National Institute of Standards and Technology. AI Risk Management Framework (AI RMF 1.0). NIST AI 100-1. Gaithersburg, January 2023. https://www.nist.gov/itl/ai-risk-management-framework
  26. Reinsurance News. "Testudo launches AI insurance underwriting platform backed by Lloyd's Lab." https://www.reinsurancene.ws/testudo-launches-ai-insurance-underwriting-platform-backed-by-lloyds-lab/
  27. ResultSense. "London insurers cap cyber payouts on AI and LLMjacking losses." 22 April 2026. https://www.resultsense.com/news/2026-04-22-insurers-cap-cyber-llmjacking-ai-payouts/
  28. Hunton Andrews Kurth. "Affirmative Artificial Intelligence Insurance Coverages Emerge." https://www.hunton.com/hunton-insurance-recovery-blog/affirmative-artificial-intelligence-insurance-coverages-emerge
  29. National Law Review. "Affirmative Artificial Intelligence Insurance Coverages Emerge." https://natlawreview.com/article/affirmative-artificial-intelligence-insurance-coverages-emerge