Read the docs

Discourse AI retries can expose private content to other users

Priority LOW 39Published · Updated

Key facts

Attack vector
Network
Privileges required
Low
User interaction
None
Attack complexity
High
Impact type
—
Confidentiality · integrity · availability
High · None · None
Fix status
Fixed

Attack profile from CVSS 3.1 · GHSA-5r2q-fr9x-wfxv. Fix status from GitHub.

Summary

Discourse AI retries run agent tools with the permissions of the original prompt author rather than the retrying user. A user allowed to use the AI bot who retries another user's prompt can read private content or copy private artifacts into a conversation they can see. Exploitation requires a shared conversation and the model selecting an affected tool. The CVSS 3.1 vector rates confidentiality impact high with low privileges required.

Written with Claude Haiku 5.5 from the sources below · 2026-11-12

Description from the GitHub advisory

### Impact A user allowed to use the AI bot could retry a response to another user's prompt and cause agent tools to run with the original prompt author's permissions. When the original author had access to private content, configured tools could retrieve that content or copy private artifacts into a conversation visible to the retrying user. Exploitation requires a shared conversation, an eligible retrying user, and the model selecting an affected tool. Unauthorized private reads and artifact imports were reproduced with request-level regression tests. Integrity impact has not been demonstrated. ### Patches A fix is prepared. Patched release versions have not yet been established. ### Workarounds Disable the AI bot to prevent new retry requests until the fix is deployed.

Why it matters now

Fix Fixed Discourse: 2026.11.0-latest.1, 2026.10.1, 2026.9.2 +1

SSVC decision points

Not available: CISA has not published SSVC decision points for this record.

Affected products

VendorProductAffected versionsFixed inStatusSource
—Discourse—2026.11.0-latest.1Patch availableGHSA-5r2q-fr9x-wfxv
—Discourse>= 2026.10.0-latest2026.10.1Patch availableGHSA-5r2q-fr9x-wfxv
—Discourse>= 2026.9.0-latest2026.9.2Patch availableGHSA-5r2q-fr9x-wfxv
—Discourse>= 2026.7.0-latest2026.7.5Patch availableGHSA-5r2q-fr9x-wfxv
  • Discourse Vendor— Affected — Fixed in 2026.11.0-latest.1 Patch available GHSA-5r2q-fr9x-wfxv
  • Discourse Vendor— Affected >= 2026.10.0-latest Fixed in 2026.10.1 Patch available GHSA-5r2q-fr9x-wfxv
  • Discourse Vendor— Affected >= 2026.9.0-latest Fixed in 2026.9.2 Patch available GHSA-5r2q-fr9x-wfxv
  • Discourse Vendor— Affected >= 2026.7.0-latest Fixed in 2026.7.5 Patch available GHSA-5r2q-fr9x-wfxv

Remediation

Upgrade Discourse to 2026.7.5, 2026.9.2, 2026.10.1, or 2026.11.0-latest.1, matching the release line in use. Until the fix is deployed, disable the AI bot to prevent new retry requests.

Written with Claude Haiku 5.5 from the sources below · 2026-11-12

  • Fix · GHSA-5r2q-fr9x-wfxv Discourse Fixed in 2026.11.0-latest.1 Advisory →
  • Fix · GHSA-5r2q-fr9x-wfxv Discourse Fixed in 2026.10.1 Advisory →
  • Fix · GHSA-5r2q-fr9x-wfxv Discourse Fixed in 2026.9.2 Advisory →
  • Fix · GHSA-5r2q-fr9x-wfxv Discourse Fixed in 2026.7.5 Advisory →

Timeline

  1. 2026-11-12 Threatist record created from GHSA-5r2q-fr9x-wfxv

References

Sources

This product uses the NVD API but is not endorsed or certified by the NVD. Sources and licences