Insurance AI-DLC DORA | Moring
DORA reaches an insurance AI-DLC through function, not through any mention of AI. Regulation (EU) 2022/2554 has applied to insurers and reinsurers authorized under Solvency II since January 2025, and any agent acting in production is an ICT-supported function. A governed lifecycle supplies what DORA examines: ICT risk management, logging and event reconstruction, resilience testing, third-party risk, and incident reporting.
Short answer
DORA reaches an insurance AI-DLC through function, not through any mention of AI. Regulation (EU) 2022/2554 has applied to insurers and reinsurers authorized under Solvency II since January 2025, and any agent acting in production is an ICT-supported function. A governed lifecycle supplies what DORA examines: ICT risk management, logging and event reconstruction, resilience testing, third-party risk, and incident reporting.
How it works
DORA's reach into an insurance AI-DLC is functional rather than nominal. The regulation never says AI; it says ICT-supported functions, and an agent that changes records, moves data, or touches critical infrastructure is one in every sense that matters. DORA became applicable to in-scope financial entities, including insurance and reinsurance undertakings authorized under Solvency II, on 17 January 2025, so European carriers adopting agent-driven delivery are already inside its perimeter.
The obligations that reach agent deployments are five. Compliance analysis for insurers enumerates them: ICT risk management under a documented, board-approved framework covering identification, protection, detection, response, and recovery; logging and event reconstruction; resilience testing; third-party risk; and incident reporting. An AI-DLC maps naturally onto several of these, because the lifecycle already emits per-change records, approval trails, and rollback paths that an event-reconstruction exercise would otherwise struggle to assemble.
Logging is the clearest fit. DORA-aligned guidance for AI agents lists logging and event reconstruction among the obligations that decide deployments, and a governed lifecycle that records every agent action, approval, and release produces that log as a by-product of work. The alternative, reconstructing what an agent did across ticketing systems and chat threads after an incident, is precisely the weakness incident reporting deadlines punish.
Resilience testing extends to the agents themselves. If an insurance AI-DLC is part of delivering or operating critical functions, then its own failure modes, such as an agent outage, a runaway loop, or a corrupt release, belong in resilience scenarios. Operational guidance for agents under DORA highlights timely reliable evidence, idempotency against duplicate transactions, and tested failover as baseline capabilities, and a lifecycle with failure drills treats these as tested paths rather than aspirations.
Third-party risk reaches the toolchain. Agents depend on model providers, coding platforms, and vendor governance layers, and DORA's third-party obligations require insurers to manage the resilience of those dependencies. Compliance guidance for insurers stresses that carriers remain responsible for every AI-driven decision whether built internally or sourced from a vendor, which aligns the contractual question directly with the lifecycle's evidence standards.
Incident reporting closes the loop. When an agent-driven change causes an ICT disruption, the record set the AI-DLC maintained, including who approved the change, which evidence supported it, and how it was rolled back, becomes the factual basis for the report. An insurer that cannot produce that record quickly is reporting an incident it cannot explain, which is a poor position under any supervisory review.
The design conclusion for European carriers is to configure the AI-DLC against DORA from the start. Tag changes by the functions they affect, retain logs in reconstructable form, include agent failure modes in resilience testing, extend third-party terms to the agent toolchain, and rehearse the incident path. The lifecycle then demonstrates operational resilience rather than asserting it.
Frequently asked questions
Does DORA apply to an insurance carrier's AI-DLC?
Yes, by function. DORA never mentions AI; it governs ICT-supported functions at financial entities, and insurers and reinsurers authorized under Solvency II have been in scope since 17 January 2025. An agent that changes records or touches critical systems is an ICT-supported function in every sense that matters, so its governance must survive the same scrutiny as any other part of the carrier's IT estate, which is exactly what a governed AI-DLC is designed to demonstrate.
Which DORA obligations bear most directly on an AI-DLC?
Five obligations reach agent deployments: ICT risk management under a board-approved framework, logging and event reconstruction, resilience testing, third-party risk, and incident reporting. The AI-DLC contributes most to logging, since every agent action, approval, and release is recorded as work happens, making event reconstruction a retrieval task. Resilience testing should also cover the lifecycle's own failure modes, such as agent outages and runaway loops, before they are discovered during a real incident.
How does an AI-DLC support DORA incident reporting?
By supplying the facts the report needs. When an agent-driven change causes an ICT disruption, the lifecycle record shows who approved the change, which evidence supported release, what the agent actually did, and how it was rolled back. DORA-aligned guidance emphasizes timely, reliable evidence for incident investigation and reporting, and an insurance carrier with governed records can meet reporting deadlines with facts rather than reconstructions assembled from tickets and memory.
What should European insurers configure before scaling agents under DORA?
Five things: tag changes by the ICT functions they affect, retain logs in reconstructable form, add agent failure modes to resilience testing, extend third-party risk terms to the model and tooling chain, and rehearse the incident path end to end. DORA-aligned guidance also points to idempotency and tested failover as baseline capabilities for agents. Configuring these into the AI-DLC before scale turns the regulation from a constraint into evidence the lifecycle already produces.