FastGPT updates data retention and deletion policies for enhanced transparency

FastGPT, an open-source AI application platform, has updated its public documentation to clarify how it handles the retention and deletion of various types of data. The changes are part of an effort to enhance transparency and user trust, especially in light of growing concerns over data privacy and security.

The platform now documents the retention periods for four key categories of data: conversation records, knowledge base files, model-call traces, and audit logs. Previously, this information was scattered across the privacy policy, per-version upgrade notes, and API reference pages. The new consolidation aims to make it easier for users to understand how and for how long their data is stored and eventually deleted.

For the cloud service, the privacy policy explicitly states that data deletion initiated by a user is a physical deletion and is not recoverable. Non-physical deletions, if they occur, will be noted in the service. The policy also confirms that user data is not stored as additional backup copies and is not used for model training. The last update to this page was recorded on March 3, 2024.

Model-call traces, which are used for short-term debugging, are retained for six hours by default. This period can be adjusted using the `LLM_REQUEST_TRACKING_RETENTION_HOURS` configuration variable. Suspended agent sandboxes are archived after a period of inactivity, determined by the `AGENT_SANDBOX_ARCHIVE_INACTIVE_DAYS` variable, with a default of seven days. In contrast, audit logs have been moved to cold archive storage upon expiry in version 4.16.0, as traceability is prioritized over prompt removal.

Three additional boundaries are documented alongside the defaults. The API endpoint responsible for clearing conversations affects only those created through an API key and does not impact conversations initiated through the web or shared links. Automatic cleanup, which is handled by background tasks, can sometimes fail. To ensure data has been properly deleted, users should verify both the request and the completion of the deletion. For community self-hosting and commercial private deployment, retention and cleanup are governed by the deploying organization, and the environment variables provide adjustable controls rather than a compliance conclusion.

Launched on September 3, 2026, the latest version, v4.16.2, is part of the ongoing efforts to improve data management. The GitHub repository for FastGPT, maintained by labring, currently has 29,551 stars and 7,297 forks across 275 releases.