Skip to main content
Redberry Labs policies cover the risks specific to deploying AI agents — not the generic exposures addressed by standard tech errors-and-omissions or cyber policies. Coverage is split into first-party losses (costs your organisation bears directly) and third-party liabilities (claims brought against you by others). Each coverage area is assessed and priced individually for every agent you schedule onto your policy.

First-party coverage

Covers direct losses your organisation incurs — including data restoration, business interruption, incident response, infrastructure overage, and incorrect funds transfer.

Third-party coverage

Covers liabilities arising from claims by third parties — including harmful or infringing AI outputs, misrepresentation, misselling, and regulatory breach.

​
Coverage areas

Coverage areaTypeWhat it covers
Data and model restoration and retrainingFirst-partyCosts to restore corrupted data or retrain a model following an AI-related incident
Business interruptionFirst-partyLost revenue and increased operating costs resulting from an AI system outage or failure
Incident response and regulatory costsFirst-partyForensic investigation, legal counsel, regulatory notification, and crisis communications costs following an incident
Runaway usage and infrastructure overageFirst-partyUnexpected compute or API costs caused by an agent behaving outside its intended parameters
Incorrect funds transferFirst-partyFinancial losses resulting from an AI agent initiating or approving a payment in error
Harmful or infringing AI outputsThird-partyClaims from third parties alleging harm caused by AI-generated content, including intellectual property infringement
Misrepresentation and missellingThird-partyClaims arising from an AI agent providing inaccurate information that a third party relied on to their detriment
Regulatory breachThird-partyFines, penalties, and defence costs arising from a regulator’s investigation into AI-related conduct

​
How coverage is priced

Coverage is priced at the agent level, not the organisation level. Each agent you submit for evaluation receives its own premium, which is determined by:
  • Deployment context — the environment the agent operates in and the sensitivity of that environment
  • Permissions and data access — the scope of what the agent can read, write, call, or execute
  • Tools — the external systems and APIs the agent can interact with
  • Users — whether the agent serves internal staff, business customers, or members of the public
  • Volume — the number of interactions or transactions the agent processes
  • Guardrails and controls — the strength of the permissioning, human-in-the-loop gates, auditability, and model governance measures you have in place
Agents with strong controls and narrow permissions attract lower premiums. Agents with broad access, financial authority, or high-volume public-facing deployment attract higher premiums.
You have a duty of disclosure. You must provide accurate and complete information about each agent at the time of submission and notify Redberry Labs promptly if an agent’s configuration changes materially. Failure to disclose may result in a claim being reduced or declined.
Before submitting your agents, review the agent risk assessment guide to understand what controls Redberry Labs looks for and how they affect your risk score and premium.