Insurance Control Plane Compliance | Moring
Compliance in an insurance control plane means the layer's enforcement and records satisfy what regulators can request: the NAIC program's documentation expectations, the EU AI Act's high-risk record requirements, and DORA's operational resilience obligations. Governance is the mechanism; compliance is the demonstrated result. A control plane makes the result producible by operation rather than assembly.
Short answer
Compliance in an insurance control plane means the layer's enforcement and records satisfy what regulators can request: the NAIC program's documentation expectations, the EU AI Act's high-risk record requirements, and DORA's operational resilience obligations. Governance is the mechanism; compliance is the demonstrated result. A control plane makes the result producible by operation rather than assembly.
How it works
Governance and compliance are related but distinct, and conflating them produces either checklists where operational control is needed or frameworks with no regulatory grounding. Compliance is what external authorities require: laws, regulations, and supervisory expectations that specify what the carrier must do with AI systems. Governance is the program of policies, controls, and accountability that makes compliance possible and sustainable. A control plane is where that program becomes executable, so its compliance value lies in producing the records and enforcing the rules the obligations actually cite.
For insurers, the anchor obligation is the NAIC Model Bulletin, adopted in December 2023 and taken up state by state. It expects a written AI program covering governance, risk management, internal controls, third-party oversight, and documentation regulators may request in an examination, and its concern is preventing inaccurate, arbitrary, capricious, or unfairly discriminatory outcomes for consumers. A control plane maps to that structure directly: the inventory and ownership records satisfy program documentation; enforcement on tool calls produces the policy decision records; approval gates evidence human oversight; and exception and escalation entries complete the evidence set internal audit can test.
The EU AI Act adds scope for carriers operating in European markets. It classifies certain insurance uses, including life and health risk assessment and pricing, as high-risk, attaching expectations for risk management, transparency, human oversight, and records of the operation of the system. A control plane supplies the operational half: per-action records, approval evidence, and version references that show which system configuration produced each decision. Compliance here is inseparable from the enforcement layer, because records produced after the fact, reconstructed from memory or partial logs, rarely withstand scrutiny.
DORA reaches carriers as financial entities subject to its operational resilience regime since January 2025. Its relevance to the control plane concerns ICT risk management, resilience testing, and oversight of critical third-party ICT providers. Agents operating inside carrier workflows are part of that ICT estate, so the control plane's records of what agents did, under which dependencies, and with which third-party services feed both incident response and the register of ICT third-party arrangements the regulation contemplates.
Operationalizing this is a control-mapping exercise. For each obligation, identify the control plane capability that satisfies it and the record that evidences it: the bulletin's vendor oversight maps to third-party dependency records and contract terms; its consumer-outcome concern maps to approval gates and enforcement logs; the AI Act's operational records map to the action chain with version references; DORA's third-party register maps to the plane's dependency inventory. Keeping that mapping current as obligations evolve is the ongoing work, and it belongs to compliance rather than engineering.
Failure modes: treating the bulletin as satisfied by a policy document while enforcement records are absent, which fails the first examination that requests evidence; mapping obligations once and never revisiting the mapping as state adoptions and AI Act guidance evolve; and leaving third-party agents outside the plane entirely, so vendor-powered decisions generate no records the carrier can produce.
A carrier starting here should build the obligation-to-control mapping as a living document, verify one obligation end to end by pulling its evidence from the plane, and repeat the verification each cycle. Compliance demonstrated from operational records is durable; compliance asserted from documents is a standing finding.
Frequently asked questions
How does a control plane map to the NAIC Model Bulletin?
The bulletin expects a written AI program and documentation regulators may request in an examination, with the stated goal of preventing inaccurate, arbitrary, capricious, or unfairly discriminatory consumer outcomes. The control plane supplies the operational half: the agent inventory and ownership records satisfy program documentation, enforcement on tool calls produces the policy decision records, approval gates evidence human oversight, and exception and escalation entries show who was notified and what was decided. Mapping each bulletin element to a plane capability and a record keeps the program demonstrable.
What changes when an insurer's AI falls in the EU AI Act's high-risk scope?
The record burden rises from good practice to obligation. Systems the regulation classifies as high-risk, including life and health risk assessment and pricing, carry expectations for risk management, transparency, human oversight, and records of the operation of the system. A control plane addresses the operational requirements by capturing per-action records with version references and approval evidence, produced continuously rather than reconstructed. Carriers should identify which of their agent workflows fall in scope and verify those specifically, since partial coverage leaves the in-scope decisions unevidenced.
Why is DORA relevant to an insurance control plane?
DORA treats insurers as financial entities with operational resilience obligations, applying since January 2025, covering ICT risk management, resilience testing, and oversight of critical third-party ICT providers. Agents acting inside carrier workflows are part of that ICT estate: their dependencies on external model providers and data services belong in the third-party register, and their records of operation feed incident response. The control plane centralizes exactly those records, which lets the resilience function see agent dependencies alongside the rest of the ICT landscape.
How should compliance teams maintain the obligation-to-control mapping?
As a living document reviewed each cycle, not a one-time matrix. For each obligation, record which control plane capability satisfies it and which stored record evidences it, then verify by pulling that evidence from the plane without engineering help. Revisit the mapping when state adoptions of the bulletin change, when AI Act guidance evolves, and when the plane's own capabilities change. Unmapped obligations are gaps by definition, and the mapping is what makes their discovery routine rather than examination-driven.