Zero Data Retention Architecture

The Core Privacy Guarantee

The defining security principle of ARMES is Zero Data Retention (ZDR) at the inference layer.

Traditional consumer AI interfaces store every prompt, attachment, and response in provider databases for 30+ days or longer. Many providers reserve the right to review transcripts with human contractors or use your conversations to train future foundation models and monetize user data.

In ARMES, every single model request is routed through cryptographic and contractually enforced Zero Data Retention infrastructure.

armes.ai

Expand

4-Scope ZDR Enforcement

ARMES routes all model calls through enterprise infrastructure managed by OpenRouter with zero-data-retention parameters strictly enforced on every payload. Under our enterprise configuration, OpenRouter enforces a comprehensive 4-Scope Zero Data Retention Policy:

[ User Prompt ] 
      │ 
      ▼ (Encrypted in transit TLS 1.3)
[ ARMES Secure Proxy ]
      │ 
      ▼ (Zero-Retention Enforced Payload)
[ OpenRouter Enterprise Infrastructure ] 
      │ ├── 1. No prompt logging
      │ └── 2. Mandatory upstream zero-retention compliance
      │ 
      ▼ (Direct API Call)
[ AI Provider (OpenAI, Anthropic, Google, DeepSeek, xAI, etc) ]
      │ ├── Ephemeral RAM processing only
      │ └── Immediate purge upon token stream completion

The 4 Scopes:

1. Zero Provider Retention: Frontier AI labs (OpenAI, Anthropic, Google, xAI, etc.) process incoming tokens entirely in volatile RAM and immediately discard them once the completion stream terminates.
2. No Model Training: Your data is never used to train, tune, or evaluate any base model, checkpoint, or reward model by ARMES or any third-party lab.
3. No Human Review Queues: Your transcripts are exempt from third-party human safety review or content moderation queues.
4. No Ad Profiling: Your queries are never mapped to advertising identifiers, cross-site cookies, or behavioral marketing graphs.

Dual-Layer Privacy Invariant

To understand your data boundaries, remember the Brain vs. Vault distinction:

External Cloud Storage Boundaries

When you connect third-party cloud services like Google Drive:

1. Google Terms Apply: Documents in your Google Drive reside on Google servers governed by Google's privacy terms. ARMES does not store these files in the Vault unless you explicitly choose to import them.
2. Encrypted Token Broker: ARMES encrypts your OAuth credentials with AES-256-GCM and never exposes raw credentials to AI models. Agents interact with external files through an ownership-scoped broker.
3. Prompt Injection Quarantine: Text read from external documents is automatically wrapped in <external_data> quarantine tags and capped at ~30k tokens to protect LLM context windows against prompt-injection attacks.
4. Action Proposal Gateway: Agents can draft modifications, but dangerous overwrites are intercepted and require your explicit interactive confirmation before any external file is altered.
5. Full Auditability: Review all agent reads, searches, and edits in the Integration Activity & Audit Log in Settings.

Verification & Auditing

You can audit data handling at any time:

• Inspect our live terms at armes.ai/privacy and armes.ai/security.
• Model routing logs show transparent timestamps and token tallies without persisting message content.