AI underwriting without governance is just automated risk.
Insurers run agents for claims, underwriting, fraud detection and customer service. TapPass keeps policyholder data contained, puts a person on the decisions that need one, and keeps every underwriting decision and claims assessment auditable.
✓Free for one person · No credit card needed
Agents read claims and score risk. They also touch life events and health records.
Three agents insurers run today, and what TapPass watches on each.
Policyholder data is redacted before the model, and payouts over your limit wait for a claims handler.
Every decision is on a hash-chained record with the data classes it used, ready to explain.
Cancellations and policy changes wait for a person, and every customer-facing action is recorded.
Why insurance AI needs governance now. The EU AI Act names it.
The EU AI Act specifically targets automated insurance decisions, and your agents handle health records, financial histories and life events.
Underwriting rules, for agents. Written by the people who own the risk.
Claims, data protection, fraud and actuarial teams each say what worries them. TapPass makes it a rule.
Runtime governance for insurance agents. From first notice to payout.
Full-stack governance for every agent in your insurance operations.
Solvency II and the EU AI Act, answerable.
What each rule asks of you, and the part of the answer TapPass gives you.
| Regulation | What it asks | What TapPass gives you |
|---|---|---|
| Solvency IISystem of governance | Sound governance and risk management of the systems that drive your decisions. | A documented, replayable trail of what each agent decided, on which data, and who approved it. |
| EU AI Act Art. 12Record keeping | Automatic logs for high-risk AI, which includes AI used in insurance decisions. | A hash-chained record of every underwriting decision, claims assessment and risk score. |
| EU AI Act Art. 14Human oversight | People able to oversee and override AI in insurance. | Approval gates on payouts, cancellations and underwriting decisions, signed by named people. |
| GDPR Art. 9Special category data | The highest protection for health data in claims and underwriting. | Health data redacted or tokenised before any model, and kept with EU-hosted providers. |
| IDDInsurance Distribution Directive | Acting in the customer’s best interest in distribution and service. | Service and sales agents work inside the rules you write, and every customer-facing action is recorded. |
No feature guarantees a compliance outcome. TapPass gives you the controls and the evidence that make your own answer defensible.
Proof you can check. Not logos you have to trust.
We do not put customer names on a page unless we can show you the work. Here is what you can verify yourself, on your own agents.
Connect one agent in watching mode. TapPass records what it would have allowed, held, redacted or blocked, and nothing is blocked while you learn.
The same rule and the same call give the same verdict, every time. Replay a decision for your auditor, or a draft rule over last week’s traffic.
Security controls, subprocessors, and a DPA on request. Hosted in the EU, built in Belgium.
What we can show you today is the product on your own agents, the documentation behind every control in the Trust Center, and a team that answers your security questionnaire itself.
Can a person review an underwriting rejection?
Yes. You decide which decisions wait for a person, such as rejections, cancellations or claims above a threshold. The reviewer sees the decision, the data classes used and the rule that fired, and signs with a passkey.
How do you spot drift in a risk-scoring agent?
TapPass watches statistical signals such as the distribution of classifications, cost and tool use per agent, and runs canary tests. A change after a model update is flagged before it shows in your loss ratios.
What stops a fraudulent claim from steering the agent?
Incoming documents are scanned for injected instructions before the agent reads them, and their content is tainted as untrusted, so it cannot trigger a payout on its own.
Can health data stay out of life pricing?
Yes. A rule on data class and agent blocks a pricing agent from reading health data, and every attempt is on the record.
Underwrite AI risk the way you underwrite everything else.
Connect one agent and see every call TapPass would have held, stopped or routed elsewhere. Nothing is blocked while you learn.
✓ EU hosted✓ Built in Belgium✓ Your own model keys✓ Encrypted at rest and in transit✓ DPA on request