"Data that doesn't exist cannot be subpoenaed, leaked, or stolen."
1. Ephemeral First
ARGUS-Thesis operates on a strict Session-Scoped Ephemeral Mode. Your inputs, claims, and generated critiques exist only in volatile RAM memory. When the node process ends or your session times out (42 minutes), the memory is zeroed.
2. Zero Persistence
We do not maintain a database of user submissions. We do not write log files containing your intellectual property. We do not use your data to train our own models. Your data exists only for the millisecond duration of the network request to the LLM provider.
3. Cryptographic Deletion
Upon clicking "End Session", the system performs a memory cleanup and generates a Deletion Certificate. This is a timestamped hash proving the destruction of the session's temporary keys and data structures.
4. Third Party Processors
To perform the audit, snippets of your text are sent to Large Language Model providers (e.g., Google Gemini, OpenAI). These providers are subject to their own stringent Enterprise Data Privacy policies, which generally preclude training on API-submitted data.