Data Privacy & Retention
How AI providers handle your data
Viro does not store prompt or response content in its own systems. Requests are processed by the model provider you select, and each provider follows its own retention policy. Paid API content is not used to train general-purpose models by default. We disclose the differences below so you can choose the right privacy level for your application.
| Provider | Viro privacy label |
|---|---|
| Viro | ZDR Verified |
| TensorX | ZDR Verified |
| Nscale | ZDR Verified |
| xAI / Grok | 30-Day Safety Retention |
| Anthropic / Claude | No Default Retention |
| OpenAI | 30-Day Safety Retention |
| Google / Gemini | 55-Day Safety Retention |
Retention labels apply to prompt and response content. Viro and its providers may retain limited operational metadata, such as token counts, model names, costs, timestamps, and request IDs, for billing, security, and reliability.
- *Claude conversation content is generally not retained by default, but this is distinct from a formal ZDR arrangement. Claude Fable 5 and Claude Mythos 5 require 30-day retention.
- Stateful features—including stored files, batch jobs, conversation history, prompt caching, and external tools—may have separate retention policies.
- Provider policies may change. Viro reviews these policies and updates its privacy labels accordingly.
Last reviewed: August 1, 2026.