For insurers, brokers and risk managers
Telematics for AI conduct.
Motor became priceable when verified telemetry existed. AI conduct has no telemetry, so when a loss arrives the only record of what the AI did is the defendant’s own logs. This is that missing record: content-free, tamper-evident, and verifiable by a party adverse to everyone involved.
The problem
The AI liability line is being built without its raw material.
Every priced line began with a loss history. This one has none, so the market is pricing on model testing and on questionnaires, while exclusions are being written at the same time as the cover is being asked for. In professional lines the losses have already arrived, and the underwriting question is already on the renewal form. What is missing is any way to answer it except with a promise.
7 of 13
Lawyers’ professional-liability carriers reporting actual AI-related claims, and adding AI-governance questions at underwriting.
EPIC 16th LPL survey, 2026
$80,000+
Average defence cost in a malpractice matter, before any judgment is entered.
Verified in pricing research
May not cover
Standard malpractice policies may not respond to AI-related claims at all. The coverage gap is the anxiety.
ABA Journal
The precedent
Cyber underwriting went through this. Early policies took the insured’s word on controls, and the word turned out to be worth very little at claim. The market moved to evidence of controls as a condition of cover, and priced accordingly. AI has arrived at the same point without the equivalent evidence existing yet.
The case that makes it concrete
In Johnson v. Dunn the firm had an AI policy in place and its lawyers were sanctioned anyway. The policy was not the failure. The failure was that nothing existed to show whether it had been followed on the day.
The four uses
What a carrier gets, what it replaces, and who can check it
| What a carrier gets | What it replaces | Who can check it |
|---|---|---|
| Self-evaluating coverage conditions. A condition written against attested operating state, checkable continuously by both sides, with neither party seeing the other’s content | A questionnaire answered once at bind and never tested again | Insurer and insured, independently |
| The renewal answer, pre-formatted. An annual governance attestation shaped as the answer to the E&O renewal questionnaire, generated locally from records already made | A partner’s best recollection, typed into a box in October | The broker, at renewal |
| Firm-level roll-up, verified contribution. An aggregator verifies every signature and rejects any that fail, producing an organisation-level view while the raw records never leave the machines that made them | Self-reported survey answers, or nothing at all | The organisation, and every contributor |
| Post-incident attribution. Reconstruct from records alone whose authority the act ran under, whether a human adopted the output as their own, and whether anyone had a window to disown it and let it pass | Depositions, reconstruction and argument | A party adverse to everyone involved |
The first of these runs today. The rest is what we want to build with underwriters rather than at them, which is the whole point of the conversation.
The plainest version
A driving record, for AI.
Conduct you can price. A sealed, content-free record of how AI is actually used, verifiable without trusting the insured, where the gaps in coverage are themselves recorded. A reinsurer can audit AI use across a whole book without opening a single document, because there is no document to open.
October is renewal season. A firm that can hand its broker a signed, contemporaneous governance record instead of an assurance is a different risk on the same form.
The ask
Bring us a book, and we will show you the record behind it.
We are looking for underwriters and claims specialists in professional lines, cyber and technology E&O to answer one question: which fields of a conduct record would actually move an underwriting or a claims decision. Nothing here asks you to trust us. Every artifact verifies offline, against a public key, with no contact with us at all.
Bring us a book, and we will show you the record behind it.
Nothing here asks you to trust us. Every artifact verifies offline, against a public key, with no contact with us at all.