Security at Conclavik

All API traffic is encrypted with TLS 1.2+ (HTTPS). Connections using older protocols are rejected. HSTS is enforced with preload. Custom security configurations are available for enterprise clients on request.

Sensitive job data (questions, analysis results, and full reports) is encrypted with AES-256 before being written to the database. Encryption keys are stored separately from the data.

All analytical content (questions, analysis, reports) is automatically purged after 90 days by default. Cryptographic attestation records, containing only content hashes, timestamps, and processing parameters, are retained indefinitely to support your compliance and audit requirements. These attestation records contain no question text, analysis content, or personally identifiable information. You can delete individual runs at any time, or request complete account erasure.

For maximum privacy, enable ephemeral mode on any request. Results are available for retrieval, then automatically purged within one hour. A tamper-evident attestation record (containing only cryptographic hashes and processing parameters) is retained to support your compliance and audit requirements.

User authentication via enterprise SSO (Google, GitHub, email). API keys are generated with cryptographically secure random generators. Authentication and credential management is handled by a dedicated identity provider. Session tokens have configurable expiration.

Runs on dedicated infrastructure hosted in Germany (EU). No shared cloud, no multi-tenant architecture. Data at rest is encrypted. Access restricted to authorized personnel only.

API endpoints are protected by per-key rate limiting. Each user's data is fully isolated; you can only access your own jobs and results. Admin endpoints require a separate admin key.

Every completed analysis generates a signed attestation certificate, a cryptographic proof that a specific analysis was performed at a specific time using specific models. Content hashes, digital signatures, model version fingerprints, and timestamps are retained indefinitely. Your compliance team can verify any report's authenticity via our public verification endpoint, without Conclavik needing to retain the report itself. You control where reports are stored. We prove they're genuine.

Every analysis uses the top-ranked model from each of multiple distinct AI providers, selected via independent public benchmarks. Models are version-pinned and recorded in your attestation certificate. If a model is unavailable, it is substituted with the next-highest-ranked alternative from a different provider, never silently. All substitutions are logged in your attestation certificate. Benchmarks are refreshed daily. Full policy available on request.

Each client's data is strictly isolated by API key. There is no cross-client data access, no shared state between analyses, and no aggregation of query patterns across users. Conclavik does not track which tickers, sectors, or securities you research. AI models are accessed via stateless API calls; they do not retain context between requests. No employee accesses client queries in normal operations; all database access requires explicit justification and is logged. Full information barrier policy available on request.

Full right to erasure: delete individual runs, purge all job content, or delete your entire account and all associated data permanently. Data portability: download each analysis result in JSON or PDF format via the API.