Book a demo
Topics · GOVERNANCE-AGENTS

Insurance Agent Registry Dora | Moring

DORA applies to AI agents at insurers by function, not by name: any agent supporting a critical or important insurance function falls within the ICT risk framework of Articles 5 to 16 for in-scope entities. The registry maps which agents carry DORA obligations and anchors the evidence, resilience, and incident-response expectations the regulation demands from 17 January 2025.

On this page
  1. Short answer
  2. How it works
  3. FAQs
  4. Related reading
  5. Sources

Short answer

DORA applies to AI agents at insurers by function, not by name: any agent supporting a critical or important insurance function falls within the ICT risk framework of Articles 5 to 16 for in-scope entities. The registry maps which agents carry DORA obligations and anchors the evidence, resilience, and incident-response expectations the regulation demands from 17 January 2025.

How it works

DORA, Regulation (EU) 2022/2554, became applicable to in-scope financial entities, including in-scope insurance and reinsurance undertakings, on 17 January 2025. Trussed's compliance guide is precise about scope: DORA's text does not reference AI or agents explicitly, but its obligations are function-based, so any system supporting a critical or important insurance function, including an agent doing claims triage or automated decisioning, falls within Chapter II, Articles 5 to 16.

That means an insurer's agent registry is also a DORA instrument. The registry identifies which agents touch critical functions, records their owners and permissions, and links them to the ICT risk framework's identification, protection, detection, response, and recovery requirements. An insurer that cannot enumerate its agents cannot credibly claim to have identified the ICT risk they introduce.

Evidence is where DORA bites hardest. Cockroach Labs' analysis of DORA database requirements states that compliance depends on timely, reliable evidence for incident investigation and reporting, and identifies capabilities that support it: append-only audit storage enforced at the storage layer, per-agent identity attribution on every write, strong read consistency, and storage-layer idempotency. Per-agent attribution is only meaningful if the registry defines the agents the writes are attributed to.

VDF's framework work for regulated industries describes the operating pattern: classify each use case by risk, limit the data and models it can touch, keep a human approver where the law expects one, and log every run so an auditor can replay it, with controls mapped to the EU AI Act and DORA and enforced at runtime. The registry is the index that makes crosswalks maintainable: each registered agent points to its tier, controls, and evidence.

Incident response completes the loop. When an agent supporting a critical function misbehaves, DORA's detection and response expectations apply, and the registry accelerates containment by identifying the agent's owner, permissions, version, and downstream dependencies. Credo AI poses the dependency question directly: when one agent fails, do you know which others break. For a in-scope insurer, that answer is part of operational resilience, not curiosity.

The practical posture for an insurer: mark DORA-relevant agents in the registry by function, apply the Articles 5 to 16 framework to them explicitly, retain durable per-agent evidence, test failover including agent dependencies, and report incidents through the registry-linked chain of ownership. Treating the registry as DORA infrastructure, rather than documentation, is what makes the compliance real.

Frequently asked questions

Does DORA apply to AI agents at insurance companies?

Yes, by function. Trussed's guidance explains DORA never names AI, but any system supporting a critical or important insurance function, including agents doing claims triage or automated decisioning, falls within its ICT risk framework for in-scope insurers. The regulation has applied since 17 January 2025, and its Articles 5 to 16 require documented, board-approved identification, protection, detection, response, and recovery for those systems.

How does an agent registry support DORA compliance?

It maps which agents support critical functions and anchors the evidence DORA expects. Cockroach Labs notes compliance depends on timely, reliable evidence for incident investigation, supported by append-only audit storage and per-agent identity attribution on every write. The registry defines those agents, their owners, permissions, and versions, and links each to its controls, making the ICT risk framework's identification and protection requirements demonstrable rather than asserted.

What evidence should insurers retain for DORA-relevant agents?

Durable, per-agent records: approvals linked to versions, decision traces, tool calls, retrieved data, access grants and revocations, and consumption records, retained in append-only storage with strong consistency. Cockroach Labs identifies these database capabilities as supporting DORA's evidence demands. The registry indexes the records by agent and workflow, so an incident investigation can reconstruct what a claims or underwriting agent did without assembling logs from scattered systems.

How does the registry help when a critical insurance agent fails?

Immediate containment and dependency analysis. The registry identifies the agent's owner, current version, permissions, and which other agents or workflows depend on it, which is the answer to Credo AI's question: when one agent fails, do you know which others break. For a in-scope insurer, that capability is part of DORA's detection and response expectations for critical functions, and it shortens both remediation and the incident report that follows.

Sources

  1. cockroachlabs.com
  2. moring.ai
  3. trussed.ai
  4. dataiku.com
  5. gatekeeperhq.com
  6. vdf.ai
  7. - content.naic.org
  8. - artificialintelligenceact.eu
  9. - eur-lex.europa.eu