Delegated authority does not delegate liability: the exam reaches you, not your MGA.
What is already moving
A sector built on delegated trust
has no record of its AI use.
The exam reaches you, not your MGA
Delegated authority, live now
When a carrier delegates underwriting authority to an MGA or coverholder, the work moves but the oversight obligation stays with the carrier. Regulators are explicit that AI governance responsibility extends to AI a delegate runs on your behalf. Binder due diligence has always checked premium, claims and bordereaux quality. It has never asked to see a delegate's AI governance program, or where a human signs off.
FCA on accountability and AI
Nikhil Rathi, FCA
As AI moves from supporting decisions to making them, accountability has to stay clear. The regulator's direction of travel is explicit: firms will be judged on how fast they surface problems once they know something has gone wrong, not on whether a policy document existed.
Consumer Duty needs evidence
FCA, ongoing
Fair value and good outcomes have to be evidenced, not asserted. If AI plays any part in pricing, underwriting screening or claims triage, the firm needs to be able to show how that output was reviewed, not just that a human theoretically could have looked.
The same shift, in the US
NAIC, pilot through Sep 2026
The NAIC's AI Systems Evaluation Tool is being piloted across a dozen US states, with formal adoption expected at the Fall 2026 National Meeting. The NAIC's AI Model Bulletin has already been adopted by more than half of US states. Governing AI use in insurance is shifting from principle to an actual exam item on both sides of the Atlantic.
The EU named insurance pricing directly
EU AI Act, deadline moved to 2 Dec 2027
Annex III of the EU AI Act classifies AI used for risk assessment and pricing in life and health insurance as high risk, by name, not by inference. The Digital Omnibus agreement pushed the compliance deadline from August 2026 to December 2027. Sixteen extra months is not the same as no obligation. The system named in the regulation is still the system named in the regulation, and a governance record built now is worth more than one built the month the new date arrives.
PI renewal questionnaires
Already changing
Renewal questionnaires increasingly ask about AI controls directly. A timestamped review trail is the difference between a standard premium and an uncomfortable conversation with your own insurer.
Your own marketing
Active now
Broker and MGA websites are advertising. Unsubstantiated claims, comparison rates and urgency wording are regulated the same way for insurance intermediaries as for anyone else.
Cryptographically sealed. Not just stored.
Evidence a carrier
would actually accept.
Every review, sealed the moment it happens.
Each check and each sign off is written into a hash chain using SHA-256. Edit, delete or backdate any record and the break is provable. Nobody has to take your word for it, including a carrier auditing your delegated authority.
Independent timestamps a carrier can weigh.
High value records are sealed with an RFC 3161 trusted timestamp from an independent authority. Anyone can verify it with standard tools, without trusting our database or yours.
Verify a record yourself, right now.
We publish a live verification page. Open a real audit record, check the chain, confirm nothing was altered. No account, no sales call. If you would not accept unverifiable evidence from a delegate, do not accept it from software either.
A report you can hand to anyone.
Every check produces a PDF with the score, the flags, the reviewer and the timestamp. Built to be handed to a carrier, an insurer, a compliance file or a regulator without a covering explanation.
Specific, not vague
The rules we check
broker and MGA marketing against.
Delegated authority oversight
A carrier delegating underwriting authority remains responsible for the AI governance of whoever holds that authority, including MGAs and coverholders.
What we check
A sealed governance record naming who owns the AI decision, what it's allowed to do, and when that authority expires or needs renewal.
FCA Consumer Duty
Fair value and good outcomes have to be evidenced when AI plays any part in pricing, underwriting or claims screening.
What we check
Named reviewer, timestamped decision, sealed record, not a policy document nobody can point to.
NAIC AI Model Bulletin
Adopted by more than half of US states. Requires a documented AI governance program with accountable ownership, for firms with a US book.
What we check
Same underlying evidence, jurisdiction mapped, so one governance record answers both sides of the Atlantic.
EU AI Act, Annex III
Names life and health insurance risk assessment and pricing directly as high risk AI. The compliance deadline moved to 2 December 2027, the obligation did not disappear.
What we check
A sealed record naming who owns the pricing or risk model, what it is allowed to do, ready well ahead of the new date rather than the month it arrives.
CAP Code / ASA
Broker and MGA marketing claims, comparison rates and urgency wording are regulated advertising, the same as any other sector.
What we check
Fake discount patterns, manufactured urgency, testimonial and comparison claims, the core of our 30 risk categories.
UK GDPR and PECR
Marketing emails need a lawful basis, honest sender identification and a working opt out.
What we check
Email compliance checks on outbound marketing copy.
And the honest scope line: we are not your MGA, your carrier or your reinsurer. We cover the published word and the AI governance record, the two places where the evidence either exists or it does not.
Inside the binder
Four desks. One record.
MGA and coverholder principals
The delegated authority is only as strong as the file behind it.
Underwriters are already using AI screening tools, with or without a governance record. The question a carrier will ask is whether you can prove oversight when it matters. This turns an unmanaged risk into a documented, supervised process.
Compliance and risk officers
Your file needs evidence, not assurances.
When a carrier or the FCA asks how AI is controlled, the answer needs dates, names and records. The governance assessment maps your gaps across 6 dimensions, and every subsequent check builds the evidence file for you.
Underwriters
AI risk is now a binder due diligence item, or it will be.
Binder reviews have always checked premium, claims and bordereaux quality. A sealed AI governance record is the difference between a routine renewal and a delegated authority that gets pulled.
Marketing and BD teams
Broker websites and client alerts are regulated advertising.
Check any page or client communication against 30 risk categories across 11 jurisdictions in under 60 seconds, with every flag explained in plain English before it publishes.
Start with the free checks. The evidence file builds itself from there.
The difference
“We reviewed it” is a claim.
A sealed record is a fact.
Without a record
✕AI oversight happens in email threads and memory
✕No timestamp, no named reviewer
✕The carrier asks, you reconstruct
✕Renewal questionnaire arrives, you hope
✕One incident, no defence
With Red Flag AI Pro
✓Every check logged with reviewer and timestamp
✓Records sealed into a SHA-256 hash chain
✓RFC 3161 trusted timestamps on key events
✓PDF evidence ready for a carrier, insurer or regulator
✓One incident, instant answer
Know where you stand in 5 minutes
Two free checks.
Both halves of the risk.
The governance assessment scores your AI oversight across 6 dimensions and shows the single biggest gap. The compliance check reads any page or document against 30 risk categories across 11 jurisdictions in under 60 seconds. Free, no account, results instantly.
Rolling this out across your delegated authorities, with client workspaces, white label reports and a managed audit trail? Talk to us about Sentinel, scoped and priced to your book.