FastGPT Centralizes Data Retention Policies for AI Workflows

The Hangzhou-based platform has moved its retention documentation out of fragmented policy pages and version notes, providing a unified look at how four distinct data classes are handled. For cloud users, the policy confirms that manual deletion triggers an immediate, unrecoverable purge, with no secondary backups maintained for model training purposes.

Technical configurations now define the lifecycle of transient data. Model-call traces default to a six-hour window, while agent sandboxes are archived after seven days of inactivity. Audit logs follow a distinct path: rather than deletion, expired logs are shifted to cold storage to ensure long-term traceability. Users should note that automatic cleanup depends on background processes, which may occasionally fail; the platform advises verifying deletion requests independently rather than assuming completion. For organizations self-hosting the software, these retention parameters are controlled via environment variables, placing the responsibility for compliance directly on the deploying entity.