Secure Enterprise AI Platform | Unified Access to GPT-4o, Grok, DeepSeek & More

Data Privacy by Design: Minimized Exposure, Maximum Control

Our AI Platform is built with privacy and security at its core. All AI queries are transmitted through API calls within our private cloud infrastructure—never through public consumer-facing endpoints. This ensures that your prompts and responses stay protected, isolated, and never exposed to public training pipelines by default.

We connect to third-party AI models (such as OpenAI, xAI, DeepSeek, Meta, etc.) exclusively through API-level integrations managed by OpenRouter, a secure routing platform that gives us full control over data handling. OpenRouter ensures:

• No prompt or response logging by default — unless explicitly enabled.
• Only anonymous metadata collection (e.g., token count, latency).
• Model-level control — we can block routing to any model whose policy does not meet our data privacy standards.

Learn more about OpenRouter’s data practices:

• OpenRouter Privacy & API Routing Overview
• Prompt Logging Settings on OpenRouter

How Each Model Provider Handles Your Data via API

Below is a provider-by-provider summary of how data is handled based on official documentation:

1. Microsoft Azure OpenAI Service

• Prompts and completions are not used for training.
• No long-term storage of user data.
• Azure OpenAI Data Zones and Updates

2. OpenAI API (via OpenAI directly)

• No training use of API data by default (as of March 1, 2023).
• Abuse monitoring data may be retained up to 30 days.
•  OpenAI Data Controls
•  OpenAI Usage Policies

3. xAI: Grok (via OpenRouter or xAI)

• Inputs are not used for training unless explicitly agreed upon.
• Some retention for abuse prevention may occur.
•  xAI Privacy Policy
• xAI Terms of Service
•  xAI Enterprise Terms

4. DeepSeek AI (via OpenRouter)

When accessed through OpenRouter, prompts and responses are not retained or used for model training, unless prompt logging is explicitly enabled by the user.

• OpenRouter only collects anonymous metadata (e.g., token count, latency) for analytics and billing.
• Learn more: OpenRouter Privacy Practices

Note: If DeepSeek is accessed directly via its own API (not through OpenRouter), user inputs, outputs, and chat logs may be used for model training, according to DeepSeek’s own terms.

5. Meta LLaMA (via OpenRouter)

• API calls to Meta LLaMA models via OpenRouter are private by default — prompts and completions are not stored or used for training unless logging is enabled.
• OpenRouter handles only anonymized usage stats for operational purposes.
• While Meta’s Acceptable Use Policy governs how LLaMA models should be used, all data privacy and retention are managed by OpenRouter in this setup.
• Learn more: Meta: LLaMA 3.2 1B on OpenRouter, Meta Responsible Use Guide

Note: Meta’s own Acceptable Use Policy governs how the LLaMA models can be used, but OpenRouter controls data handling when acting as the intermediary.