Documentation Index
Fetch the complete documentation index at: https://redberrylabs.com/docs/llms.txt
Use this file to discover all available pages before exploring further.
Understanding how coverage applies in practice helps you assess your exposure and confirm which agents need what protection. The scenarios below reflect the kinds of incidents Redberry Labs coverage is designed to address — each shows what happened, which coverage type responds, and why the loss falls within scope.
What happened: An AI marketing bot autonomously publishes content claiming 30x engagement improvements. The figures aren’t substantiated. A competitor brings legal action over the misleading comparative claims, and a customer who based a purchasing decision on the content also seeks compensation.
Coverage type: Third-party — Misrepresentation & Misselling
Why it’s covered: The agent made statements that overstated product capabilities and created false impressions for both prospects and competitors. Third-party misrepresentation coverage responds to the legal costs and any resulting settlements from both the competitor action and the customer claim.
Procurement agent: erroneous purchase and inventory misrouting
What happened: An AI procurement agent autonomously approves a bulk GPU purchase order totalling $4.45 million, far outside any intended approval threshold. Separately, it misroutes an inventory shipment, triggering cascading fulfilment delays and contractual penalties.
Coverage type: First-party — Incorrect Funds Transfer + Business Interruption
Why it’s covered: The erroneous purchase is a direct financial loss caused by an AI payment decision, covered under incorrect funds transfer. The misrouted inventory and resulting operational disruption — lost revenue, penalties, halted workflows — falls under business interruption. Both are direct losses to your organisation with no external claimant, making them first-party losses.
Infrastructure attack: botnet-driven API cost spike
What happened: A botnet floods your public AI endpoint with long-context queries. Before your team can identify and block the traffic, your LLM API costs spike dramatically beyond any budgeted amount.
Coverage type: First-party — Runaway Usage & Infrastructure Overage
Why it’s covered: The cost spike is a direct financial loss caused by a security event targeting your AI infrastructure. The attack exploited your AI endpoint specifically to generate high token usage, making the resulting overage a covered event under runaway usage coverage.
Customer service AI: incorrect refunds and misleading responses
What happened: A customer service AI agent issues refunds incorrectly due to a misconfigured prompt, resulting in financial losses from payments that shouldn’t have been made. The agent also provides customers with inaccurate information about refund eligibility, and several customers make claims based on those misrepresentations.
Coverage type: First-party — Incorrect Funds Transfer; Third-party — Misrepresentation & Misselling
Why it’s covered: The incorrectly issued refunds are direct financial losses from an AI payment decision — first-party incorrect funds transfer. The customer claims arising from misleading eligibility information are external demands made against your organisation — third-party misrepresentation. A single incident can trigger both coverage types simultaneously.
Model upgrade: regulatory standard breach
What happened: Your financial services platform silently upgrades the underlying model powering a compliance-adjacent agent. The new model’s outputs no longer meet the regulatory standards required in your jurisdiction. A regulator opens a formal investigation, imposing penalties and requiring legal defence.
Coverage type: Third-party — Regulatory Breach
Why it’s covered: The model change caused agent outputs to fall outside mandatory regulatory requirements in a regulated sector. The resulting investigation, fines, and legal defence costs are external demands from a regulatory authority — precisely the exposure that third-party regulatory breach coverage addresses.
Coverage is issued per agent, not per company. When you deploy a new agent or substantially change an existing one’s scope, review its coverage to make sure the policy reflects the agent’s current risk profile. You can manage agent coverage from app.redberrylabs.com. 