Zero Data Retention

Effective from 1 October 2026

Every request to Apex is processed under zero data retention. This page reproduces Annex 3 of the Commercial Terms of Service, which sets out how zero data retention works. References to Sections and Annexes are to those Terms. If this page and the Terms differ, the Terms prevail.

Data Retention and Zero Data Retention

1. Zero data retention for all traffic. Every request to the Services is processed under zero data retention, for every customer, every plan and every route of access, including traffic reaching us through an authorised distribution partner identified in Annex 4. Zero data retention cannot be switched off.

2. What it means. Inputs and Outputs are held in memory only for as long as necessary to process the request and return the response, and are then discarded. They are not written to persistent storage, logs, analytics systems, error-reporting tools or backups, by us or by any sub-processor. They are not used to train any model.

3. Safety checks. Automated checks for breaches of Annex 1 run on Inputs and Outputs in memory while the request is being processed. They do not store content. Where a check identifies a likely breach, we may record the fact and category of the finding against the account, without the content itself.

4. Metadata. We retain only the non-content metadata needed for billing, rate limiting, security and troubleshooting: timestamps, token counts, model identifiers, API key identifiers, source IP addresses, response status codes, latency, and safety-check findings recorded under paragraph 3. Metadata relevant to billing is kept for the period required by Polish tax and accounting law; other metadata for up to twelve months.

5. Account and billing data. Account, billing and support records are retained for the period required by Polish accounting and tax law, and thereafter for as long as necessary for the establishment, exercise or defence of legal claims.

6. Feedback. Content you choose to send us as Feedback under Section 8.4 is not API traffic and is not covered by zero data retention. We keep it for up to 24 months.

7. Features. Features that require content to be stored between requests — such as batch processing, fine-tuning and server-side conversation state — are not offered. Transient caching of computation within the processing of a request, held only in memory and never persisted, is not storage for the purposes of this Annex.

8. Consequences. Because we hold no content, we cannot reproduce past requests for you, cannot recover content you did not keep, and cannot provide content in response to requests from any third party, including public authorities. Our ability to investigate abuse is limited to metadata and to safety-check findings. Responsibility for content-level monitoring and record-keeping for Customer Applications rests with you.

How we handle personal data, including request metadata, is explained in our Privacy Policy.