Verisk, Armilla Shift Liability Coverage as AI Legal Errors Surge in 2026
Verisk and Armilla AI are updating legal malpractice policies as high-profile AI-driven legal errors mount in 2026.
Why it matters: Legal professionals increasingly use generative AI, but unintentional outputs ('hallucinations': false or fabricated content generated by AI) create new liability risks. Insurers are tightening or expanding coverage, making it essential for law firms to understand updated policy terms and how to mitigate technology-driven malpractice exposure.
- Verisk will introduce three new general liability endorsements in 2026 to let insurers exclude damages caused by generative AI outputs (CG 40 47, CG 40 48, CG 35 08).
- Armilla AI, a Lloyd’s of London coverholder, now offers $25 million liability protection for AI-specific risks, including hallucination and regulatory non-compliance.
- AI-related lawsuits in the U.S. rose 978% from 2021 to 2025 as insurers and law firms face new exposures.
- Recent examples: Sullivan & Cromwell’s court filing with false AI citations, and an attorney reprimanded in April 2026 for fabricated case law.
Insurers and law firms are scrambling to address a surge in professional liability risks caused by generative artificial intelligence. With AI tools producing 'hallucinations'—content that is inaccurate or entirely fabricated—legal professionals now face mounting scrutiny and potential malpractice claims.
- Verisk’s January 2026 general liability endorsements (CG 40 47, CG 40 48, and CG 35 08) empower insurers to specifically exclude coverage for damages or legal malpractice triggered by AI-generated outputs.
- Armilla AI, a Lloyd’s of London coverholder, now markets standalone AI Liability Policies. They cover up to $25 million for issues ranging from inaccurate information to outright hallucinations or regulatory violations. As Armilla notes, this coverage directly addresses AI-driven mistakes increasingly hitting courts and clients.
Hallucinations—AI-generated fabrications—have triggered real legal fallout. Examples:
- In April 2026, Sullivan & Cromwell revealed its team submitted court filings containing AI-fabricated case citations, earning swift rebuke.
- U.S. Magistrate Judge Robert T. Numbers reprimanded another attorney the same month, after their AI-generated filing included made-up case law.
- Deloitte’s October 2025 report to the Australian government cited academic sources invented by AI, resulting in a partial refund for flawed work.
"Some carriers are introducing policy riders, or specialized AI liability coverage, particularly for practices or systems that are relying very heavily on AI tools," said Deepika Srivastava, COO at The Doctors Company.
The bottom line: Courts are holding lawyers accountable for content they file—regardless of its origin. As AI-driven lawsuits spike (up 978% from 2021 to 2025), legal insurers and practitioners must watch for rapid policy changes, evaluate their tech risk protocols, and stay alert for new exclusions or coverage products shaped for the AI era.
By the numbers:
- $25 million — Maximum AI liability coverage per organization offered by Armilla AI’s policy.
- 978% — Increase in U.S. AI-related lawsuits from 2021 to 2025, per legal analytics.
Yes, but: While some insurers are excluding AI-driven risks, standalone liability policies are simultaneously expanding, offering new options for law firms that actively manage their AI exposures.
What's next: Verisk’s new AI-related liability endorsements take effect January 1, 2026; law firms and in-house counsel should review coverage ahead of renewals.