redberrylabs.com

What It Is

Redberry Labs positions itself as “AI Liability Insurance for Autonomous Agents.” Its core offering is not firewalls, EDR, or vulnerability scanning, but liability insurance and risk assessment for AI agents that can take actions on behalf of a company. Its view is that traditional tech E&O and cyber insurance were not designed for systems whose behavior can change based on prompts, policies, and real-time context, and therefore may not adequately cover agent-driven failures.

Core Capabilities and Cybersecurity Relevance

In terms of protection type, it is closer to AI risk transfer and insurance-based governance than direct technical defense. Redberry Labs emphasizes underwriting at the individual agent level rather than covering the company as a whole. It assesses risk based on each agent’s permissions, tools, policies/prompts, guardrails, and data access surface. The text gives examples such as procurement agents, accounts payable agents, and logistics agents. Once these agents have authority over payments, approvals, contracts, or scheduling, they can indeed introduce risks similar to business fraud, privilege abuse, compliance failures, and operational incidents.

Pricing, Deployment, and Integrations

On pricing, the company only discloses that it assesses and prices risk per agent. It does not publish premiums, plans, minimum coverage amounts, or deductibles. The deployment model is also unclear: it does not state whether access to APIs, logs, policy files, prompt repositories, or runtime monitoring systems is required. Management and alerting capabilities, incident response workflows, and claims evidence requirements are not explained either. As for integrations, the text only mentions that the company evaluates an agent’s tools, permissions, and policies/prompting; it does not disclose specific integrations with cloud platforms, identity systems, SIEM, GRC, or AI observability platforms.

Pros and Cons

The main advantage is that the entry point is both forward-looking and specific: using the actual action surface of an AI agent as the unit of risk is more granular than broad “AI company insurance.” The team’s background spans AI research, Tesla Insurance, LexisNexis, Zurich, and underwriting for cyber and professional liability insurance, giving it a foundation for building insurance products. The downside is that the public materials are still largely conceptual and team-focused, with limited information on policy wording, exclusions, regulatory qualifications, underwriting partners, applicable jurisdictions, or claims examples. Enterprises should conduct thorough due diligence before purchasing.

Who It’s For and Access from China

It is better suited to overseas enterprises or AI companies already piloting AI agents in high-privilege workflows such as procurement, finance, and logistics—especially scenarios where they need to explain to customers, boards, or compliance teams “who is responsible when AI fails.” Access from China, payment methods, and local purchase availability are not covered in the text, so they should be considered unknown. If launched in the Chinese market, enterprises would generally also need to factor in local cybersecurity, data compliance, AI governance, and insurance regulatory requirements, while considering domestic insurers, GRC platforms, or AI security assessment services as alternatives or complements.