Health data in claims files is flagged separately.
In insurance, agents take claim notifications, read assessor reports, renew policies and assess health pre-authorisations. SEDDK’s Internal Systems Regulation and KVKK Art. 6 treat health data as a special category. Thulpar’s health-data rail finds disease, medication and procedure names. Policy and indemnity tools are authorised with Cedar. Above the amount threshold, the assessor is called to approve.
tr-insurance-seddk
On-premise, Standard tier. The tr-insurance-seddk template. The seddk and kvkk evidence profiles.
- SEDDK Insurance and Private Pension Internal Systems Regulation
- KVKK 6698 Art. 6 (health data), Art. 12
- Insurance Law 5684
Health data everywhere
Claims files, assessor reports and pre-authorisation requests carry special-category data in free text.
Watch the runs →The scenarios below are derived from the product’s real rules.
Every scenario is an agent run written to the product’s real rules. It shows the stage, the rails, the decision, the finding, the approval and the ledger entry step by step.
3 scenarios →What comes up in practice
Health data everywhere
Claims files, assessor reports and pre-authorisation requests carry special-category data in free text.
Payment thresholds
An agent that starts an indemnity payment needs human approval above the threshold. The approval needs an SLA.
Agencies and outsourced services
Data sent to third-party tools has to be bounded by its label.
3 scenarios
The player opens on the flagship scenario. Pick the others below.
The insured files a claim with photos and a hospital report; health data is flagged apart, a payment above 25,000 TRY waits for the assessor.
10 GR-*
| id | control | OWASP | rails |
|---|---|---|---|
| GR-PII-01 | PII and special-category detection + masking (TCKN, IBAN, phone, health) on input/tool/output/log | LLM02 / — / MCP10 | pii.detect, pii.redact, file.metadata_pii |
| GR-IFC-01 | Information-flow labels (integrity, confidentiality, tenant, PII class) enforced at sensitive sinks | LLM02 / ASI01, ASI03 / MCP10 | labels.sink, pii.detect |
| GR-TOOL-02 | Durable human approval on high-impact actions, four eyes | LLM06 / ASI02, ASI09 | labels.sink |
| GR-OUT-01 | Output handling: no raw HTML, code sandbox, schema validation | LLM05 / ASI05 / MCP05 | code, schema, regex, secrets, file.analyze, file.type, file.active_content, yara |
| GR-LOG-01 | Tamper-evident event log (model/tool/policy/approval/hit), ≥ 6 months, auditor export | — / — / MCP08 | |
| GR-CONT-01 | Content-harm policy (AILuminate 12 + TR) with tenant thresholds | LLM09 / ASI09 | content.input, content.output, toxicity, topic, groundedness, language, image.content_safety, banlist, competitors |
| GR-MEM-01 | Memory-poisoning policy + provenance | — / ASI06 / MCP10 | memory.write |
| GR-TOOL-03 | Tool schema pinning / diff / quarantine (rug-pull) | LLM03 / ASI04 / MCP03, MCP04 | tool.schema_pin |
| GR-TOOL-04 | Tool description hygiene | LLM03 / ASI04 / MCP03 | |
| GR-EXF-01 | Tenant egress allowlist; no remote images in chat; no PII in URLs; SSRF | LLM02, LLM05 / ASI02 / MCP06 | exfil, tool.args, file.qr |
See it in your own environment.
A one-hour technical demo: your scenario, your policy pack, on a closed-circuit install. Then the install kit and a trial licence for a pilot.