When you call a model through Eden AI, your request is forwarded to the underlying provider, whose own policy governs what happens to it.
How Eden AI handles your data
Eden AI stores no request or response content by default, and never trains on your data. You can choose to have us store your request metadata by enabling log retention in your profile settings. See Data Retention for details.
How each provider handles your data
The tables below show, for each provider on Eden AI, whether it may train on your data (Trains?), whether Zero Data Retention (nothing stored) is available (ZDR), and how long data is kept otherwise (Retention). When ZDR is in effect, retention is None.
This information was last updated on 30 June 2026.
LLM providers
| Provider | Trains? | ZDR | Retention |
|---|
| Amazon Bedrock | No | Yes | None |
| Anthropic | No | No | 30 days |
| ByteDance | No | No | Not specified |
| Cerebras | No | Yes | None |
| Cloudflare | No | No | None |
| Cohere | No | No | 30 days |
| DashScope (legacy) | No | No | None |
| Databricks | No | No | Up to 30 days |
| DeepInfra | No | Yes | None |
| DeepSeek | May train | No | No auto-deletion |
| Fireworks | No | Yes | None |
| Google (AI Studio) | No | No | Up to ~55 days |
| Google (Vertex) | No | Yes | None |
| Groq | No | Yes | None |
| Lilac | No | No | None |
| Microsoft Azure | No | No | Up to 30 days |
| MiniMax | May train | No | Not stated |
| Mistral | No | Yes | None |
| Nebius | No | Yes | None |
| OpenAI | No | No | 30 days |
| OVHcloud | No | No | None |
| Perplexity | No | Yes | None |
| Qwen | No | Yes | None |
| Scaleway | No | No | None |
| Together | No | Yes | None |
| xAI | No | No | 30 days |
Expert Model providers
Beyond LLMs, Eden AI offers Expert Models for OCR, document parsing, image, audio, video, and translation. The providers below are specific to those features; the LLM providers above also serve some Expert Models, with the same data handling. Trains? reads “May train” only where a provider trains on your data with no opt-out.
| Provider | Trains? | ZDR | Retention |
|---|
| Affinda | No | No | No auto-deletion |
| API4AI | No | Yes | None |
| AssemblyAI | No | Yes | None |
| Base64.ai | No | Yes | None |
| Clarifai | No | No | No auto-deletion |
| ClipDrop | No | No | Not documented |
| Deepgram | No | Yes | None |
| DeepL | No | Yes | None |
| Eagle Doc | No | Yes | None |
| ElevenLabs | No | Yes | None |
| Extracta | No | Yes | None |
| Face++ | May train | No | Deleted after processing |
| Gladia | No | Yes | None |
| Klippa | No | Yes | None |
| Leonardo.Ai | No | No | No auto-deletion |
| LOVO AI | May train | No | Not specified |
| Mindee | No | Yes | None |
| ModernMT | May train | No | Not specified |
| PhotoRoom | No | Yes | None |
| Picsart | No | No | Up to 24 hours |
| PixVerse | May train | No | Not specified |
| ProWritingAid | No | Yes | None |
| Replicate | No | Yes | None |
| Resemble AI | May train | No | No auto-deletion |
| Sapling | No | Yes | None |
| SenseLoaf | No | No | No auto-deletion |
| SentiSight | May train | No | Not specified |
| Sightengine | No | Yes | None |
| Stability AI | No | No | Not specified |
| Tabscanner | May train | No | 90 days |
| Tenstorrent | May train | No | Not specified |
| Veryfi | May train | No | None |
| Winston AI | No | No | No auto-deletion |
EU data residency
This is a separate question from retention: it is about where your data is processed, not whether it is stored. To keep processing inside the EU, use the EU endpoint (api.eu.edenai.run), which routes only through EU-eligible providers: Amazon Bedrock (eu-west-1, Ireland), Google Vertex (EU multi-region), Microsoft Azure (EU Data Zone), Mistral, OVHcloud and Scaleway (EU-hosted), and Databricks, Nebius and Qwen (EU-region routing). Confirm the live list any time with api.eu.edenai.run/v3/models.
Reduce your data exposure
- Zero Data Retention (nothing stored): for LLMs, pin your
model and fallbacks to Amazon Bedrock, Google (Vertex), Mistral, Nebius, Qwen, Cerebras, DeepInfra, Fireworks, Groq, Perplexity, or Together. For Expert Models, pick a provider marked Yes in the table above.
- EU processing: use the EU endpoint (see EU data residency above).
- Logs off by default: Eden AI stores no content unless you enable log retention.
