CVE-2026-43533: OpenClaw: QQBot media tags could read arbitrary local files through reply text
OpenClaw before 2026.4.10 lets QQBot media tags in AI reply text point at any local path on the host. The outbound media handler read files outside the media storage boundary and sent them out, which discloses local files. Upgrade openclaw to 2026.4.10 or newer.
Published Apr 17, 2026 · Updated May 5, 2026 · Analysis Oct 29, 2026
What an attacker can do
The QQBot channel lets an AI reply carry media tags that name a file to send as an image or attachment. In affected versions, a tag could reference a host-local path that lies outside the intended media storage directory, and the outbound handler would read it and deliver it to the chat. This is path traversal (CWE-22) through model-written text.
The attacker does not need shell access. They need to influence what the AI replies on a QQBot channel, either as a low-privileged user chatting with the agent or through prompt injection in content the agent reads, and ask it to emit a media tag with a path such as a config file, key file or environment file. The file's contents arrive in the chat the attacker can see. The CVSS 4.0 vector rates confidentiality impact as high and the rest as none. Whatever the OpenClaw process can read is at risk: API keys, tokens, SSH keys and source code.
Who is affected
The openclaw npm package before 2026.4.10, with the QQBot channel enabled. Check the version:
npm ls -g openclaw
openclaw --versionInstances that do not use QQBot do not run the outbound media path described in the advisory.
openclaw
- Ecosystem
- npm
- Affected
- < 2026.4.10
- Fixed in
- 2026.4.10
- Check
npm ls openclaw
How to fix it
Upgrade to 2026.4.10 or newer. Release 2026.4.14 includes the fix.
npm install -g [email protected]Pull request 63271 (commit 604777e4414cc3b2ff8861f18f4fb04374c702c6) enforces the media storage boundary on every outbound QQBot local file path. The advisory names no workaround. If you suspect disclosure, rotate secrets stored in files that the OpenClaw process can read.
How to tell if you were hit
Search QQBot reply logs and chat transcripts for media tags whose path points outside the media directory, such as absolute paths or ones with parent-directory segments, and for file attachments that you did not expect. Beam records file access by agents on the host, so reads of sensitive files by the OpenClaw process around a reply would show up there.
Original advisory
From the GitHub Advisory Database (CC BY 4.0).
Summary
QQBot media tags could read arbitrary local files through reply text.
Affected Packages / Versions
- Package:
openclaw - Ecosystem: npm
- Affected versions:
< 2026.4.10 - Patched versions:
>= 2026.4.10
Impact
QQBot outbound media tags in AI reply text could reference host-local paths outside the intended media storage boundary, allowing local file disclosure through outbound media handling.
Technical Details
The fix enforces the media storage boundary for all outbound QQBot local file paths.
Fix
The issue was fixed in #63271. The first stable tag containing the fix is v2026.4.10, and [email protected] includes the fix.
Fix Commit(s)
604777e4414cc3b2ff8861f18f4fb04374c702c6- PR: #63271
Release Process Note
Users should upgrade to openclaw 2026.4.10 or newer. The latest npm release, 2026.4.14, already includes the fix.
Credits
Thanks to @feiyang666 of Tencent zhuque Lab (https://github.com/Tencent/AI-Infra-Guard) for reporting this issue.
Credited to feiyang666.
Questions
- What can an attacker read with CVE-2026-43533?
Any local file the OpenClaw process can read, by getting the AI reply to include a QQBot media tag with that path. The advisory describes it as local file disclosure.
- Is a shell or code execution possible?
Not according to the advisory. The impact is confidentiality only, per the CVSS vector.
- Which version fixes it?
2026.4.10 or newer. The fix keeps all outbound QQBot local file paths inside the media storage boundary.
- Does it matter if I don't use QQBot?
The advisory concerns QQBot outbound media tags, so deployments without that channel do not use the affected path.
References
How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology
More OpenClaw vulnerabilities
All OpenClaw entriesCVE-2026-62197
OpenClaw: Browser CDP discovery could accept blocked WebSocket URLs
medium8.5Oct 21, 2026CVE-2026-35630
OpenClaw: QQBot native approval buttons did not enforce configured approver identity
high8.0Jul 2, 2026CVE-2026-53806
OpenClaw: Combined POSIX shell options could confuse exec revalidation
high8.8Jul 2, 2026CVE-2026-53809
OpenClaw: Embedded runner policy could be confused by provider aliases
medium3.8Jul 2, 2026CVE-2026-53810
OpenClaw's marketplace runtime extension metadata could point at unscanned payloads
high8.8Jul 2, 2026CVE-2026-53811
OpenClaw: Matrix allowFrom could bind to mutable display names
high8.8Jul 2, 2026
