Security, compliance, and data handling for TapPass. Built in Belgium, deployed in the EU.
In preparation. Controls implemented:
In transit: TLS 1.2+ on all connections. Internal services use mTLS.
At rest: Credential vault AES-256-GCM. Audit integrity SHA-256 hash chain.
Every LLM call passes through a multi-step pipeline: PII detection, prompt injection scanning, secret detection, exfiltration prevention, output scanning, and human approval gates.
Deploy in your infrastructure. Your data never leaves your network. We have no access to your instance.
Hosted in the EU. Request content processed in memory, not persisted. Audit metadata stored in EU.
| Data | Stored | Details |
|---|---|---|
| Request/response content | No | Processed in memory. Forwarded to LLM. Not persisted. |
| Audit metadata | Yes | Timestamp, agent ID, classification, cost. Hash-chained. |
| Credentials | Yes | AES-256-GCM encrypted. API keys stored as hashes. |
| Configuration | Yes | Pipeline settings and policies. No PII. |
Stateless API gateway with policy engine sidecar. All decisions computed per-request, no shared state between replicas. Horizontal scaling.
Every customer is a separate organization. All queries, audit entries, credentials, and policies scoped by org_id. No cross-org API access.
SDK retries with backoff. Agents fail closed by default.
RetryAll requests blocked until recovery.
Fail closedAuto-fallback to configured backup provider.
FallbackAudit writes locally. Pipeline continues. Syncs on recovery.
Local fallbackManaged service only. Self-hosted deployments have no sub-processor dependencies.
| Name | Role | Location |
|---|---|---|
| Supabase | Database and auth | EU (Frankfurt) |
| GitHub | Source code and CI/CD | US (source code only) |
| Cloudflare | CDN and DDoS | Global (TLS termination only) |
LLM providers (OpenAI, Anthropic, Azure) are your sub-processors, not ours. TapPass proxies requests using your API keys.
Unit, integration, and security test suites. SAST, SCA, secret, and container scanning in CI on every commit.
Internal red team testing across all detection categories including adversarial prompt injection and exfiltration attempts.
Scheduled. Report available under NDA upon completion.
Automated via governance pipeline. PII exposure, injection attempts, and secret leaks create tracked issues with auto-escalation.
Configurable per severity: webhook, Slack, email, SIEM. Breach notification within 72 hours per GDPR Art. 33.
Report vulnerabilities to security@tappass.ai. Acknowledged within 24 hours.
| Tier | Response | Channels | Hours |
|---|---|---|---|
| Community | Best effort | GitHub Issues | Business hours CET |
| Professional | 4h P1, 1 day P2 | Email, Slack | Business hours CET |
| Enterprise | 1h P1 | Dedicated Slack | 24/7 for P1 |