Redberry Labs
Liability insurance for autonomous AI agents, priced per agent by simulating failure scenarios since there's no historical claims data to underwrite against.
What it is
Redberry Labs is a London-based managing general agent (MGA) that underwrites, prices, and sells liability insurance for autonomous AI agents, paying claims off a reinsurer's balance sheet. Founded by Dr. Charles Higgins and Dr. Sophia Kalanovska, both AI PhDs from King's College London who previously built and sold Tromero, a GPU aggregation platform (£1.5M seed led by BlueYard Capital). Since agentic AI has no loss history to price against, Redberry runs an underwriting engine that uses AI itself as the simulation layer, playing an agent against thousands of scenarios to estimate the probability and cost of failure. Pricing is per agent based on what it can touch — its 'blast radius' — rather than per company, and coverage spans first-party losses (business interruption, model retraining) and third-party claims (an agent's action harming someone else).
Who it's for
Founders running AI agents past Higgins's autonomy line — agents that send money, make bookings, or dispatch shipments rather than just recommending — who are discovering their general liability, cyber, and tech E&O policies were never written to respond to an AI agent's own bad judgment call.
Why it's better
- •Prices risk by simulation instead of requiring historical claims data that doesn't exist yet for agentic AI failure modes.
- •Prices per agent on blast radius — what systems, money, and people it can reach — so scoping down an agent's permissions directly lowers the premium.
- •Arrives ahead of the market: over 80% of state requests from Berkshire Hathaway, Chubb, Travelers, and others to exclude AI from general liability were approved starting January 2026, via ISO endorsements CG 40 47, CG 40 48, and CG 35 08.
- •One line of integration gives Redberry direct visibility into an agent's behavior, or a founder can skip integration and pay more for the added uncertainty.
- •Built by outsiders to insurance on purpose — two AI safety PhDs who treated 'nobody underwrites this' as the opportunity rather than a warning sign.
Heard on AI for Founders
Redberry Labs was featured in an episode of the AI for Founders podcast. Hear the full conversation.
Listen to the EpisodeRelated Tools in Security
Enterprise AI services company built on a hybrid routing model that decides which workloads a frontier model is allowed to see, and which never leave your own environment.
Privacy, AI, and security consulting for founders who are vibe coding at midnight and need someone to tell them what's on fire before a regulator does.
AI governance and cybersecurity consultancy — privacy in America is already dead, and Optimal IT helps companies build AI tools with real guardrails using the OWASP Top 10 for LLMs.
Explore more AI tools for founders
Browse All Tools