The FTC probe puts a price on rogue-agent liability

The MGA idea is taken. Armilla, AIUC, Klaimee, Mount, Testudo, Relm and Munich Re all underwrite AI or agent liability. AIUC already ties cover to a controls and testing standard, and Klaimee and Mount sell to mid-market companies deploying agents. A new entrant would need a sharp niche, such as one vertical like fintech payment agents, or a capacity relationship the others lack. Claims forensics is still open. I found no product that reconstructs an agent incident from logs and assigns fault between the model provider, the deployer's configuration and the user's instruction in a form adjusters can settle on. What exists is general forensic consulting (J.S. Held), gateway-vendor logging content (Gravitee) and 2026 academic proposals (the arXiv AI insurance stack blueprint and the CER claim-grade evidence framework). Those papers say adjusters will need this capability and that closed claims history barely exists. The opening is to be the neutral party carriers bring in after a loss: a standard evidence format plus a defensible way to split fault, sold to the carriers and MGAs listed above, not against them. The risks are that agent claims volume is still small, and that MGAs like AIUC and Mount, which already run pre-bind testing, may build forensics in-house. One way in is to give deployers the logging standard that makes them eligible for cover, then sell the resulting incident data back to underwriters as pricing data.

• Armilla AI

  Lloyd's coverholder MGA dedicated to AI liability. In January 2026 it raised limits to $25M per organization and explicitly added autonomous-agent cover: incorrect decisions, improper tool use, escalation errors. This is the closest direct rival to the Agent Liability MGA.

• AIUC (Artificial Intelligence Underwriting Company)

  Founded by an ex-Anthropic executive; reported $40M raised. Pairs its AIUC-1 agent certification standard (quarterly adversarial testing, independent audits) with Lloyd's-backed insurance, which is essentially controls as policy conditions. Reportedly backs a policy for ElevenLabs voice agents.

• Klaimee

  YC Spring 2026 company that bills itself as a liability insurer for autonomous AI agents. Offers risk evaluation, certification, financial guarantees and liability insurance; reported $5.5M raise. Founder previously structured a Tokio Marine MGA.

• Mount

  YC company calling itself an AI-agent insurance carrier. First product is AI liability cover for mid-market and enterprise firms deploying autonomous agents, priced after scanning and red-teaming the deployment.

• Testudo

  Lloyd's coverholder selling generative AI third-party liability cover. Capacity raised to $9.25M per insured with Atrium and QBE in 2026.

• Relm Insurance (NOVAAI, PONTAAI, RESCAAI)

  Specialty carrier with three AI products: tech E&O for AI platforms, an excess wrap for third-party AI liability, and first-party response cover for companies that embed third-party AI.

• Munich Re aiSure (and Mosaic partnership)

  Big-carrier AI performance-error cover for AI providers and corporate adopters. A 2026 Mosaic partnership offers up to 15M in initial limits to AI developers and vendors.

• J.S. Held

  Established forensic consulting firm with an AI-in-claims practice. Traces losses through AI outputs, human reliance and system handoffs, and advises preserving logs, prompts, approvals and configurations. This is the nearest existing competitor to Agent Claims Forensics, but it is services, not product.

• Gravitee

  Agent and LLM gateway vendor that publishes on agent forensic logging, chain of custody and incident root-cause analysis. Could move into forensic evidence, but has no insurer-facing claims product.

Checked 2026-10-17