Executive brief
Cherry Studio, an AI productivity desktop application, contains a flaw in how it handles stored information (memory). Due to an error in how data is uniquely identified, information saved by one user or assistant could potentially block or interfere with the data of another. This could lead to unauthorized access or manipulation of data across different user profiles or AI agents.
Technical details
A vulnerability in CherryHQ cherry-studio (up to version 1.9.7) exists within the `sha256` function of `MemoryService.ts` in the CherryIN Preload API. The root cause is that memory deduplication used a global SHA-256 hash of the memory text without incorporating user or assistant identifiers. This allows a remote attacker with low privileges to manipulate the 'state' argument, leading to an authorization bypass where one user's data can collide with or block another's. A pull request (#15413) has been proposed to scope these hashes by including the user_id and agent_id, though the vendor has indicated this component may be removed entirely in version 2.0.
Affected products
- CherryHQ cherry-studio up to 1.9.7
Timeline
- 2026-05-29: patched: Pull request 15413 submitted to address the hashing scope issue.
- 2026-06-29: advisory: CVE-2026-13534 published.