Zero Data Retention
Last updated: March 26, 2026
At Levante, we believe your data belongs to you. This page explains our Zero Data Retention (ZDR) commitment: how it works, what it covers, and what guarantees we provide when you use models tagged as private through Levante Platform.
Applies to private models only.Zero Data Retention guarantees apply exclusively to models tagged as “private” in Levante Platform. Models without this tag may be subject to the data retention policies of their respective providers.
1. What is Zero Data Retention
Zero Data Retention (ZDR) means that the content of your API requests — including prompts, completions, and any attached context — is never stored by the upstream AI infrastructure providers that process your requests. This guarantee applies only to models tagged as private in Levante Platform.
When you use a private model, your prompts and responses are processed in real time and discarded immediately after delivery. No logs, no copies, no training data.
2. How It Works
When you make a request to a model tagged as private through Levante Platform:
Request received
Your API request arrives at Levante. We validate your API key and route the request to the appropriate model provider.
Ephemeral processing
The request is processed within EU-sovereign infrastructure using ephemeral enclaves. The content exists only in memory during processing.
Response delivered
The model response is streamed back to you. Once delivery is complete, all content (prompt and completion) is immediately wiped from memory.
Nothing remains
No logs of prompt content, no copies of completions, no data used for model training. Only usage metadata (see Section 4) is recorded.
3. What Data Is NOT Retained
When using models tagged as private in Levante Platform, the following data is never stored by the upstream infrastructure:
4. What Data IS Retained
Levante retains usage metadata necessary for billing, rate limiting, and service operation. This metadata does not include any content from your requests or responses:
| Data | Purpose |
|---|---|
| Tokens consumed (input/output) | Billing and spend tracking |
| Model identifier | Usage analytics and routing |
| Timestamp | Audit trail and rate limiting |
| API key identifier (prefix only) | Access control and attribution |
| Response status code | Error monitoring and reliability |
For full details on how we handle your personal and usage data, please refer to our Privacy Policy.
5. EU Data Residency
All requests to models tagged as private are processed exclusively within European Union infrastructure:
Dublin
Ireland
Frankfurt
Germany
Paris
France
This ensures full compliance with GDPR Article 44 regarding international data transfers. Your data never leaves the EU during processing.
6. Which Models Support ZDR
Not all models available through Levante operate under Zero Data Retention. Only models tagged as “private” in Levante Platform are covered by ZDR guarantees. You can identify them in the following ways:
Via the API
The GET /v1/models endpoint returns a zero_data_retention field for each model. Models with true are tagged as private and covered by ZDR.
In the Dashboard
Models tagged as private are marked with a visible “private” badge in the model selector within your team settings on Levante Platform.
If Zero Data Retention is critical for your use case, ensure you only use models tagged as “private”. Models without this tag may be subject to the data retention and training policies of their respective providers.
7. Limitations and Disclaimers
Please be aware of the following:
- Intermediary role: Levante acts as an API gateway. We route your requests to upstream model providers and do not store the content of your prompts or completions on our own infrastructure.
- Provider guarantees: The Zero Data Retention commitment for private models is enforced by our upstream infrastructure providers. We select providers that contractually guarantee ZDR and EU data residency.
- Scope:ZDR applies exclusively to models tagged as “private” and only to the content of API requests and responses. Usage metadata (tokens, timestamps, model IDs) is retained as described in Section 4.
- No training:Your data is never used to train, fine-tune, or improve AI models — neither by Levante nor by our ZDR infrastructure providers.
Important: While we take every measure to ensure ZDR compliance, Levante cannot be held liable for breaches by third-party infrastructure providers. We conduct regular due diligence and maintain contractual safeguards to protect your data.
8. Contact
If you have questions about our Zero Data Retention practices or need compliance documentation for your organization, contact us at:
Email: saulgmz@claiacademy.com