CVE-2026-85093: Incorrect Authorization in cheshire-cat-ai core

The vulnerability in Cheshire Cat AI core's GET /memory/collections/{collection_id}/points endpoint arises from a failure to apply per-user filtering when retrieving episodic memory points. Authenticated attackers possessing MEMORY:READ permission can exploit this flaw to enumerate and retrieve conversation messages and personal data belonging to all users by using the offset cursor for pagination. This results in unauthorized disclosure of sensitive information across user boundaries. The affected versions are from 1.8.0 up to and including 1.9.2. The CVSS 4.0 base score is 7.1, indicating high severity with network attack vector, low attack complexity, no privileges required beyond MEMORY:READ, no user interaction, and high confidentiality impact.

Successful exploitation allows an authenticated user with MEMORY:READ permission to access all users' stored conversation messages and personal data, violating confidentiality and potentially leading to privacy breaches. There is no indication of integrity or availability impact. No known exploits are reported in the wild at this time.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, restrict MEMORY:READ permissions to trusted users only and monitor access to the affected endpoint to limit exposure.