●PocketClawAI security alerts
Sign in
vulnerabilitymedium · CVSS 5open

CVE-2026-22807

OpenClaw — prompt-injection-resistant audit log incomplete

Timeline
Disclosed2026-04-18Patch available—Open (mitigation pending)2026-05-03

Each rust dot is a disclosed event in this advisory's life: when it was published, when (if ever) a patch shipped and where things stand today (the dashed line). Ghosted dots are events that haven't happened yet.

Summary

The OpenClaw 2026.4 audit log records tool calls but did not include the full prompt context that triggered them. A prompt-injection attack could not be reconstructed from logs alone, making forensic analysis significantly harder. Improvement scheduled but not yet shipped.

Details

CVE IDCVE-2026-22807
Severitymedium (5)
CVSS vectorAV:L/AC:H/PR:H/UI:N/S:U/C:L/I:N/A:N
Published2026-04-18
Patchednot yet
Affected versions2026.4.0 – current
Fixed inPending (planned 2026.5)
Exploited in the wildno known exploitation

Affected AI agents

OpenClaw →
Are you running OpenClaw?Paste your docker-compose.yml or requirements.txt and we'll tell you in 10 seconds whether CVE-2026-22807 hits your stack.
Scan my AI stack →

What to do

  • If you run an affected version: upgrade to Pending (planned 2026.5) immediately. Do not delay this for convenience reasons.
  • Rotate any credentials that may have been exposed via the affected component.
  • Audit your logs for indicators of exploitation — unexpected outbound traffic, anomalous tool calls, unfamiliar authenticated sessions.
  • If exploitation is confirmed, treat the host as compromised: rebuild from a clean image, rotate every secret on the host, audit lateral movement.

Sources

Are you affected?

Type the version you have installed. We check it against 2026.4.0 – current.

This is a best-effort check. When in doubt, upgrade to Pending (planned 2026.5).

Related AI CVEs

See also: all AI CVEs, AI agents tracker, scan your AI stack, Pro alerts, methodology.