Trust & Security
Authentication: The Trust-First Approach
The Brainstorm Reactor uses a dedicated authentication service as the complete login infrastructure. The decision for a dedicated, standalone sign-in/sign-up page (rather than a modal overlay) is strategic and based on the B2B positioning of the product.
Architectural Rationale
In a B2B context, trust is the largest conversion lever. A modal (pop-up) suggests fleeting B2C interaction. The dedicated page with verified trust badges provides the security signals users need in the critical decision moment — for sensitive strategic data. Additionally: clean, shareable URL (/sign-in) for IT administrators and team leaders.
Verified Trust Badges
All trust badges are based exclusively on demonstrable facts anchored in the legal documentation:
| Badge | Basis |
|---|---|
| GDPR Compliant | Processing based on Art. 6(1) GDPR / TKG 2021 |
| No Client Tracking | Exclusively server-side telemetry (PostHog Server-to-Server), no browser scripts |
| No IP Storage | IP addresses irreversibly anonymized before persistence (overwritten with 127.0.0.1) |
| AI Act Transparent | Schema.org AIAnnotations JSON-LD in <head>, ai-disclosure attributes on message nodes, C2PA/XMP metadata in PDF exports |
| Pseudonymized Analytics | Rotating 8-character hash, automatic rotation every 30 days, no re-identification possible |
| EU Infrastructure | Core infrastructure (application server, database, analytics) operated on servers in the European Union |
Infrastructure & Server Locations
EU Core (no data transfer to third countries for these services)
| Service | Provider | Location |
|---|---|---|
| Application Server | Applikationsserver | Frankfurt am Main, Germany (fra1 / eu-central-1) |
| Database & File Storage | Datenbank (PostgreSQL) | Dublin, Ireland, EU (eu-west-1) |
| Product Analytics | Produktanalyse | Frankfurt am Main, Germany (EU Cloud) |
| Live Audio Transcription | Sprachtranskription | EU Region |
US-Based Third Parties (under EU-US Data Privacy Framework / DPF, adequacy decision July 10, 2023)
| Service | Provider | Data Accessed |
|---|---|---|
| Authentication | Authentifizierung | Login tokens and session data — no access to prompts or usage data |
| Payment Processing | Zahlungsanbieter | Purely transactional payment data |
| AI Routing | AI-Gateway | Transit layer with zero-data-retention. Prompts are not stored. |
Security Architecture
Middleware Protection
Auth-Middleware protects /reactor(.*) and /api/reactor(.*) via authenticated sessions. Sign-In/Sign-Up routes are public.
Session Architecture
The Brainstorm Reactor uses URL-anchored session IDs (?session=...) — not naive global state (e.g., LocalStorage). This enables true parallel sessions across multiple browser tabs without state contamination.
AI Model Routing Privacy
All AI requests route through the AI-Gateway with:
HTTP-RefererandX-Titleheaders for transparent attribution- Zero-data-retention agreement
- No prompt storage at the routing layer
Context Pruning (Amnesie-Zwang)
Unlike ChatGPT, the Brainstorm Reactor proactively and strictly prunes the model's history. This ensures the AI in step 4 of a recipe does not hallucinate on data from step 1 — it focuses 100% on the next concrete heuristic. This is both a quality feature and a privacy protection (no accumulation of sensitive strategy data in the model's active context).
Audio Data Handling
Live audio is processed as a transient stream. Raw audio never reaches the application database. The transcription provider operates under a zero-retention agreement — audio payloads are discarded immediately after text extraction. Only the derived text artifacts (transcripts, structured insights) enter the session persistence layer, subject to the same session lifecycle rules as all other user data. No biometric identification, emotion recognition, or voice profiling takes place.