Executive brief
The mem0 server, a universal memory layer for AI agents, contains a security flaw where its memory management API endpoints do not require authentication. This allows any remote user to modify, overwrite, or delete stored memory records without permission. Such an exploit could lead to significant data loss or unauthorized manipulation of the information used by AI agents, potentially impacting their reliability and operational integrity.
Technical details
The mem0 server (version 1.0.0 and earlier) is vulnerable to a missing authentication check (CWE-306) for critical memory management API endpoints. Specifically, functions such as 'PUT /memories/{memory_id}' are exposed without any verification of the requester's identity or authorization level. A remote, unauthenticated attacker can exploit this by sending crafted HTTP requests to the network-reachable server to modify, overwrite, or delete arbitrary memory records. This results in a complete loss of integrity for the stored data. As of the advisory date, no patched version has been identified.
Affected products
- mem0ai mem0ai (mem0) <= 1.0.0
Timeline
- 2026-05-12: disclosed: Initial disclosure and NVD publication
- 2026-05-12: advisory: GitHub Advisory published