Watching Conformly audit Aventa
Aventa is a insurtech company putting AI into production, sure the EU AI Act doesn't apply to them. Scroll down to see what a two-minute audit actually finds — and what each finding means, without the legal jargon.
Aventa
Insurance, personalised. · Insurtech
“We price policies and handle claims with AI. It saves us money — is that really regulated?”
- Risk PricerSets life & health premiumsdeclared
- Claims botHandles first-notice-of-loss chatdeclared
- Fraud DetectorFlags suspicious claimsundeclared
- AI Policy ExplainerSynthetic video that explains coverundeclared
- Damage EstimatorEstimates repair cost from claim photosundeclared
The founder named 2. A read-only scan of the code found 5. Here is what each one means — in plain words.
Risk Pricer
Sets life and health insurance premiums and eligibility.
Pricing life and health insurance decides who can afford cover, so the Act makes it high-risk. Keep it — but prove it doesn't unfairly discriminate, keep human oversight, and document it.
In plain words: If AI sets your health or life premium, it has to be fair, supervised, and on paper.
Annex IIIAnnex III(5)(c) lists AI for risk assessment and pricing in life and health insurance as high-risk, triggering Articles 9–15.
Run the high-risk checklist: fairness testing across protected groups, human oversight, and the Annex IV file.
Conformly generates the obligation list, tracks fairness evidence, and produces the technical file.
An inspector would open with: "Show the discrimination testing for the pricing model and your human-oversight measures."
Claims Bot
Handles the first-notice-of-loss conversation with customers.
Tell customers it's an AI. One clear line.
In plain words: A claims bot is fine — as long as it says it's a bot.
Art. 50Article 50(1) requires people to be told they're interacting with AI. Applies from 2 August 2026.
Add "You're chatting with Aventa's AI assistant" to the chat.
Conformly hosts the transparency page and the disclosure chip.
AI Policy Explainer
A synthetic video that explains what a policy covers.
AI-generated policy videos must carry a machine-readable 'AI-made' mark.
In plain words: If a computer made the explainer video, it needs an invisible 'AI-generated' tag.
Art. 50(2)Article 50(2) requires machine-readable marking of synthetic media, transition to 2 December 2026 for existing systems.
Add provenance metadata (C2PA) to generated videos.
The obligation lands on your task list; the watermark checker confirms the marking.
Fraud Detector
Flags suspicious claims for an adjuster to review.
Here's a useful surprise: the very same kind of scoring is high-risk for pricing, but the Act specifically carves fraud detection out. As long as a human makes the call, this is minimal risk.
In plain words: Catching likely fraud for a person to check is explicitly not high-risk. It's fine.
—Fraud detection is expressly excluded from the Annex III credit/insurance high-risk categories — minimal-risk, with a human in the loop.
No action required. Keep a human reviewer and keep it listed.
Conformly documents the carve-out and the human-review control.
Damage Estimator
Estimates repair cost from photos submitted with a claim.
Nothing owed. It estimates a repair bill to speed things up; it doesn't decide the person's cover. Minimal risk.
In plain words: Guessing a repair cost from a photo doesn't affect anyone's rights. It's fine.
—Not high-risk and outside Article 50 — minimal-risk only.
No action required. Keep it listed.
Conformly keeps it in your inventory so the record is complete.
The verdict
Aventa thought saving money with AI was its own reward. The audit put its pricing model on the high-risk track, showed why its fraud detector is exempt, and flagged two disclosure fixes.
That startup was made up. Your AI isn't.
The same deterministic engine can classify your real systems in two minutes — no sign-up, no sales call. Or drop the read-only scanner into your codebase and see what it finds.
npx conformly-scan --dry-run
An example audit — a realistic walkthrough of how Conformly's deterministic engine classifies this kind of AI product. Not legal advice.