Security & Compliance
Enterprise-grade security with full compliance coverage
Tenant Isolation
InnFuse enforces strict tenant isolation at every layer of the stack. Data from one tenant is never accessible to another.
- Database Isolation Each tenant has a dedicated PostgreSQL database. No shared tables, no row-level security workarounds.
- Storage Isolation Separate object storage buckets per tenant with distinct access credentials.
- Cache Isolation Prefixed cache namespaces per tenant. No cross-tenant cache pollution possible.
- Secrets Management Per-tenant encryption keys stored in Kubernetes Secrets. Keys are never stored in the database or application config.
Encryption
InnFuse applies encryption at multiple layers to protect data at rest and in transit.
| Layer | Method | Scope |
|---|---|---|
| TLS 1.3 | TLS 1.3 | All data in transit between clients, services, and databases |
| AES-256 | AES-256 | Full disk encryption on all storage volumes |
| AES-256-GCM | AES-256-GCM | PII fields, API keys, credentials, and sensitive configuration |
EU AI Act Compliance
InnFuse implements the requirements of the EU Artificial Intelligence Act for AI systems used in business contexts.
- Transparency Users are always informed when they are interacting with an AI agent. Agent responses include clear AI disclosure markers.
- Human Oversight Human operators can monitor, intervene in, and override AI decisions at any time. All tool executions with elevated security levels require human approval.
- Data Governance All AI training and inference data is processed within the European Union. No data is transferred to third countries without adequate safeguards.
- Risk Assessment InnFuse performs automated risk classification for all AI agent configurations. Agents handling sensitive data or critical decisions are flagged for additional review.
GDPR Compliance
InnFuse fully complies with the General Data Protection Regulation (EU 2016/679). The following data subject rights are implemented.
| Right | Implementation |
|---|---|
| Right of Access | Data subjects can request a full export of their personal data via the privacy dashboard or by contacting privacy@innfuse.nl. |
| Right to Erasure | Data subjects can request deletion of their personal data. Erasure is completed within 30 days, with confirmation. |
| Right to Portability | Personal data can be exported in machine-readable format (JSON, CSV) for transfer to another service provider. |
| Right to Object | Data subjects can object to AI-based profiling or automated decision-making. Objections are processed within 72 hours. |
| Data Retention | Configurable retention periods per data category. Default: conversation data 12 months, audit logs 5 years, account data until deletion. |
Platform Audit Log
Beyond AI-specific auditing, InnFuse maintains a comprehensive platform audit log covering all administrative actions.
- Login attempts, password changes, 2FA enrollment, session management
- System settings changes, feature toggles, plan upgrades
- Agent configuration changes, model switches, prompt updates
- Data exports, deletions, and retention policy executions
- Channel connections, disconnections, and configuration changes
Audit logs are retained for a minimum of 365 days. Enterprise plans support custom retention periods up to 10 years.
AI Audit Trail
Every AI interaction is logged in an immutable audit trail for compliance and debugging purposes.
Audit logs are retained for a minimum of 5 years in accordance with regulatory requirements. They can be exported in JSON or CSV format via the API.
| Event Type | Data Logged |
|---|---|
| Tool execution | Tool name, input parameters, output, execution time, security level, agent ID |
| Knowledge base search | Query text, matched articles, relevance scores, agent ID |
| Agent handoff | Source agent, target agent, reason, conversation ID, context transferred |
| Tool approval request | Tool name, parameters, requesting agent, approver, decision, response time |
| Escalation event | Reason, source agent, target (agent or human), conversation context |
Tool Security Levels
Every tool in InnFuse is assigned a security level that determines whether human approval is required before execution.
Safe
Read-only operations with no side effects. Executed automatically without approval.
Elevated
Actions that modify external data. May require approval depending on configuration.
Dangerous
Destructive or irreversible actions. Always requires human approval before execution.