Why make TrustSignal a condition of coverage
Instead of taking an insured's AI logs on faith, every covered output carries a cryptographic receipt — sealed hash, timestamp, and fraud score — you can verify without touching any underlying PII or model content.
Underwritable risk
Every covered AI output produces a cryptographically sealed, fraud-scored receipt at generation time — converting unverifiable AI exposure into risk you can actually price.
Instant verification
Adjusters and counsel verify a receipt in seconds via API or web, confirming the evidence existed when the insured says it did and has not been altered.
Defensible denials
When receipts are missing or don't match, you can decline or right-size a claim on sealed, objective facts instead of “we don't believe your logs” — reducing bad-faith exposure.
For AI claims, your cost is skewed toward proof, not payouts.
The figures below are deliberately conservative modeled scenarios, not historical loss-ratio data. The first 6–12 month pilot is designed to replace these assumptions with your portfolio's observed leakage, severity, and loss-adjustment-expense deltas.
Modeled forensic + discovery spend to reconstruct what an AI system did and when.
A receipt-backed verification call plus minutes of adjuster time replaces weeks of forensics.
Assuming a 50% reduction in AI-related leakage at a $115k average claim size.
A pricing band you can offer while still improving combined ratio, because leakage and LAE drop.
For a carrier processing ~50 disputed AI claims a year, those operational savings compound to over $8M across the portfolio in our scenarios — before counting lower bad-faith exposure or regulatory fines. All figures are forward-looking models, validated per-portfolio during the pilot.
Low adoption friction is what makes a mandate realistic.
Requiring verification only works if it doesn't burden your insureds. TrustSignal integrates as a thin evidence layer around the AI systems they already run: a lightweight API hook for technical teams, or a web UI and proxy gateway for everyone else. The content never leaves their environment — only sealed hashes and metadata.
Insureds comply because it helps them too. Sealed evidence qualifies them for coverage or a lower premium, and it defends them against false third-party claims with the same cryptographic receipt you rely on. The product overview explains the receipt and verification lifecycle in full, and the developer documentation covers the integration path.
How to make it a condition
You don't flip the whole book on day one. The path from pilot to portfolio is staged so each step earns the next.
90-day evidence pilot
Require TrustSignal on 5–50 new or renewing policies in a high-risk AI segment. Track claim frequency, verification time, and fraud-flag rate to produce a loss-ratio or LAE proof point.
Endorsement language
Add a verified-AI endorsement to the pilot segment — first as an optional discount qualifier (lower resistance), then as a condition of coverage for the highest-risk accounts.
Portfolio rollout
Bake verification into base AI E&O policy language and launch a “verified AI” product line that competitors can't match without rebuilding their evidence stack.
Carrier questions
Won't requiring this burden my insureds and shrink my book?
Integration is a lightweight API hook around existing logs and outputs, with a web UI for non-technical accounts. Insureds benefit through better terms or qualification for coverage at all. Pilot telemetry is designed to measure adoption friction directly.
What about PII or sensitive content?
TrustSignal never sees the content. The API seals a hash of the output plus a fraud score and timestamp — not the output itself. The insured's data never leaves their environment; only cryptographic fingerprints and metadata are sealed.
What does “tamper-evident” actually mean?
Each receipt is a zero-knowledge proof over the output hash, fraud score, and timestamp. Altering any field breaks the cryptographic seal and is detectable in seconds. It cannot be forged without breaking the proof system.
What if the insured cancels mid-policy?
Policy language can require active sealing. If the verification heartbeat stops, the carrier is alerted and can trigger an audit or non-renewal. Past receipts remain verifiable permanently.
Is the fraud score a guarantee?
No. The fraud score is a risk signal (0–100), not a binary determination. It is an input to a claims decision, not the sole determinant — which keeps it defensible and avoids creating new liability.
What if TrustSignal goes away?
Receipts are self-verifying. Even if TrustSignal ceased to exist, any receipt can be verified with open-source tooling — verification does not depend on us being online.
Run a 90-day pilot on your highest-risk AI accounts.
Replace modeled savings with your actual numbers. Book a carrier briefing to review the integration path, sample receipts, and pilot structure.
Book a carrier briefing