Lloyd's coverholder Testudo expands generative AI liability capacity to $9.25M with Atrium and QBE
Testudo has added Atrium Underwriting Group and QBE Insurance Group to its (re)insurance panel, increasing available limits for its generative AI liability product to $9.25 million per insured. The new support joins Apollo Underwriting and signals a clear shift: capacity is moving to purpose-built AI cover as traditional liability policies pull back with new exclusions.
The policy is built to respond to third-party claims tied to AI-generated outputs-covering legal costs and damages where commercial general liability (CGL) may not respond. This is particularly relevant for defamation, IP, discrimination, and other harms arising from automated decisions or synthetic content.
Why it matters for insurers and legal teams
Demand is climbing as carriers introduce AI-related exclusions across CGL, E&O, and media lines. Testudo reports a 137% year-over-year rise in generative AI litigation, with scrutiny expanding across technology, financial services, healthcare, and media.
For brokers and GCs, the message is simple: assumptions about legacy policies won't hold. Definitions, exclusions, and trigger language around "publication," "professional services," "intentional acts," or "contractual liability" can leave material gaps when AI is involved.
What this product is designed to address
- Third-party claims caused by AI-generated outputs (text, code, images, video), including alleged defamation, IP infringement, and misleading or harmful content.
- Defense costs and damages where CGL and adjacent policies may be silent or excluded for AI-related perils.
- Clearer triggers around AI output liability compared to patchwork extensions or ambiguous endorsements.
Underwriting focus: controls buyers should expect to evidence
- Model governance: documented use cases, human-in-the-loop checkpoints, change management, and model/version inventories.
- Data and content risk: training data provenance, IP/license hygiene, PII/PHI safeguards, and content filtering for toxic or unsafe outputs.
- Prompt/output logging: retention, auditability, and reproducibility of decisions and generated content.
- Vendor and contract risk: indemnities, flow-down obligations, and clarity on model providers vs. deployers.
- Red-teaming and monitoring: abuse testing, jailbreak defenses, incident response, and kill-switches for faulty models or integrations.
Broker and GC checklist
- Map AI use cases by function and risk category (customer-facing content, credit decisions, clinical support, code generation, marketing assets).
- Review current exclusions across CGL, Tech E&O, Media/Advertising, and Cyber; identify overlaps and hard gaps.
- Tighten contracts with AI vendors and downstream users; negotiate indemnities and evidence of insurance aligned to AI exposures.
- Stand up logging and content review workflows; keep evidence for claims handling and regulatory inquiries.
- Align incident response with AI-specific scenarios (harmful content, bias claims, IP complaints) and panel counsel protocols.
- Seek clarity on retroactive dates, territory/jurisdiction, panel counsel options, sublimits for IP/defamation, and claims-made wording.
"Lloyd's has always led with courage when insuring new and emerging risks. We are delighted and honoured to welcome the innovative Atrium and QBE to our (re)insurance panel for our generative AI liability insurance product. Their support reflects a shared commitment to thoughtful, well-structured underwriting approaches that give enterprises greater confidence to adopt generative AI." - Mark Titmarsh, Co-Founder and Head of Insurance, Testudo.
"As a leading global insurer, it's our role to support businesses as they navigate evolving technologies and emerging risks. We're pleased to support Testudo's innovative AI liability insurance solution, helping our clients with AI risk scoring and reporting tools, to better detect and evaluate potential liabilities." - Peta Kilian, Senior Underwriter, Innovation, QBE.
Market context and next steps
With AI exclusions spreading, standalone or dedicated AI liability is becoming a practical route to ring-fence exposure and avoid disputes over intent or "publication." Expect more carriers to enter, but with stronger underwriting around governance and data rights. For buyers, the quickest wins are better documentation, tighter contracts, and transparent disclosures at bind and renewal.
Background reading that can help your underwriting submissions: the NIST AI Risk Management Framework for governance alignment (NIST AI RMF) and the Lloyd's market resources for specialty placements (Lloyd's).
Helpful resources
- AI for Insurance - training and resources on AI risk assessment and underwriting considerations.
- AI for Legal - litigation trends, compliance topics, and third-party liability issues tied to generative AI.
Your membership also unlocks:
