PatchSiren

PatchSiren cyber security CVE debrief

CVE-2026-25083 GROWI, Inc. CVE debrief

CVE-2026-25083 is a HIGH severity authorization bypass vulnerability affecting GROWI wiki platform versions 7.4.5 and earlier. The vulnerability exists in OpenAI thread/message API endpoints that fail to enforce proper access controls. A logged-in attacker with knowledge of a shared AI assistant's identifier can view and modify other users' threads and messages without authorization. The CVSS 4.0 vector indicates network attack vector with low attack complexity, low privileges required, and high impact to confidentiality and integrity. The vulnerability was published March 16, 2026 and last modified May 19, 2026. No known exploitation in ransomware campaigns has been reported.

Vendor
GROWI, Inc.
Product
GROWI
CVSS
HIGH 8.7
CISA KEV
Not listed in stored evidence
Original CVE published
2026-03-16
Original CVE updated
2026-05-19
Advisory published
2026-03-16
Advisory updated
2026-05-19

Who should care

Organizations running GROWI wiki platforms with OpenAI integration enabled, particularly those using shared AI assistants in multi-user environments. Security teams responsible for wiki/content management systems and AI/LLM integrations. Development teams maintaining GROWI deployments or similar wiki platforms with AI assistant features.

Technical summary

The vulnerability stems from missing authorization checks in GROWI's OpenAI integration API endpoints. When users interact with shared AI assistants, the thread and message APIs fail to verify that the requesting user owns or has permission to access the requested thread resources. This allows any authenticated user who obtains or guesses another user's thread identifier to read and modify conversation history. The attack requires low privileges (any logged-in user) and no user interaction, with high impact to data confidentiality and integrity. The fix involves implementing proper ownership verification on all thread and message API operations.

Defensive priority

HIGH

Recommended defensive actions

  • Upgrade GROWI to version 7.4.6 or later which contains the authorization fix
  • Review OpenAI assistant identifiers for any that may have been exposed or shared beyond intended users
  • Audit access logs for unauthorized thread/message access patterns in shared AI assistant contexts
  • Implement additional monitoring on OpenAI integration endpoints for anomalous cross-user data access
  • If immediate patching is not possible, restrict access to OpenAI assistant features to trusted administrative users only

Evidence notes

Authorization failure in OpenAI integration endpoints; CWE-862 (Missing Authorization) identified by JPCERT/CC. Affected versions confirmed as 7.4.5 and earlier per vendor advisory.

Official resources

2026-03-16