Executive brief
OpenClaw, an integration tool for communication platforms, contains a flaw in how it authorizes access to Nextcloud Talk rooms. Because the system identifies rooms by their names rather than unique internal IDs, an attacker can create a room with a similar name to bypass security filters. This could allow unauthorized users to gain access to private or protected conversation rooms, potentially exposing sensitive business discussions.
Technical details
A policy confusion vulnerability exists in OpenClaw's Nextcloud Talk integration (specifically within inbound callback and policy handling). The root cause is the reliance on user-controlled or non-unique room display names for authorization checks instead of stable, unique room tokens. An attacker with low privileges can exploit this by creating rooms with colliding names to trick the allowlist logic into granting access to protected resources. This is classified as CWE-807 (Reliance on Untrusted Inputs in a Security Decision) and CWE-639. The issue is fixed in version 2026.3.22 by ensuring authorization is keyed on stable room tokens.
Affected products
- OpenClaw openclaw < 2026.3.22
Timeline
- 2026-03-15: patched: Fix committed to repository
- 2026-03-24: advisory: Original advisory GHSA-xhq5-45pm-2gjr published
- 2026-04-09: disclosed: CVE-2026-35624 assigned
- 2026-04-18: other: Duplicate advisory GHSA-5f7h-p83x-5vc2 withdrawn
References
- https://github.com/openclaw/openclaw/security/advisories/GHSA-xhq5-45pm-2gjr
- https://github.com/openclaw/openclaw/commit/630f1479c44f78484dfa21bb407cbe6f171dac87
- https://github.com/openclaw/openclaw/commit/a47722de7e3c9cbda8d5512747ca7e3bb8f6ee66
- https://www.vulncheck.com/advisories/openclaw-policy-confusion-via-room-name-collision-in-nextcloud-talk