Lloyd's Lab - Cohort 17Illustrative loss file / eHealth
One fabricated allergy
enters the record.
Treatment changes. Months later, a claim arrives.
No penicillin allergy stated
"Penicillin allergy" inserted
Draft signed into the chart
The recorded allergy is relied upon
The event must be reconstructed
Lloyd's Lab - Cohort 17The insurance gap
At bind, supervision is a promise.
At claim, it must be a fact.
"Contradictions are checked and every note is reviewed by a clinician."
A sensible control description. No event-level proof.
- Which system and version acted?Unknown
- What authority did it have?Unknown
- Which control should have applied?Disputed
- Did it execute, and what did it decide?Unknown
- What human action followed?Incomplete
The carrier cannot distinguish the supervision before the loss or reconstruct it cleanly afterward.
Lloyd's Lab - Cohort 17Why the market needs this now
Affirmative AI cover is arriving
before affirmative proof.
CFC announced explicit AI language across products including eHealth and Technology E&O.
Public market signal - not a Glacis partnership claimEach party sees a slice. No single provider automatically preserves the whole control path.
Lloyd's Lab - Cohort 17What Glacis changes
Preserve the control path
while the AI is acting.
- Workflow
- ambient-note / 4.2
- Control profile
- contradiction-check / 2.1
- Recorded decision
- ESCALATE
- Human review
- completed / override recorded
- Integrity
- verifies
- Declared gap
- none
No raw patient content is required in the portable record.
Lloyd's Lab - Cohort 17The first insurance product
A policy schedule
the insured can satisfy.
Name the covered workflows, required controls, event scope, retention period and treatment of known gaps.
The insured produces portable records for declared events. Glacis verifies integrity. The carrier stays outside the quote path.
Either side can verify the record. Claims investigates gaps, materiality and context; the record does not decide coverage.
Underwriting can review declared coverage, exceptions and supervision history without receiving raw clinical content.
Lloyd's Lab - Cohort 17A hard, finite experiment
The Lab ends with
a stop / amend / adopt decision.
ClasseHealth / healthcare technology E&O
WorkflowAmbient clinical documentation
Population3-5 representative insured workflows
Loss testOne adversarial claims tabletop
- the requirement works without carrier quote-path integration;
- the insured can produce it without exporting raw patient content;
- outages, bypass and missing evidence are distinguishable;
- claims can answer the five event questions; and
- underwriter and broker judge the schedule placeable.
Lloyd's Lab - Cohort 17The ask
We are asking Lloyd's to make
one market decision.
- eHealth underwriterowns risk selection
- claims leadowns investigation
- wording counselowns the obligation
- brokertests placeability
- 3-5 insured workflowstest operational burden
- tested schedule wording;
- claims protocol;
- implementation and failure report; and
- commercial pilot ready for participant approval.
If AI can change the risk, proof of its supervision must exist before the claim.