PatchSiren cyber security CVE debrief
CVE-2026-65700 h2oai CVE debrief
CVE-2026-65700 is a critical path traversal vulnerability in h2oGPT's OpenAI-compatible files API. The vulnerability allows unauthenticated remote attackers to read, write, and delete arbitrary files accessible to the server process by supplying traversal sequences in the bearer token. The get_user_dir function in openai_server/backend_utils.py uses the bearer token string unsanitized as a path component via os.path.join, enabling attackers to traverse outside the intended user directory through the file content, delete, and upload endpoints to achieve remote code execution by writing to startup hooks or application-loaded files. Users of h2oGPT version 0.2.1 and prior should be aware of this critical vulnerability.
- Vendor
- h2oai
- Product
- h2ogpt
- CVSS
- CRITICAL 9.3
- CISA KEV
- Not listed in stored evidence
- Original CVE published
- 2026-07-23
- Original CVE updated
- 2026-07-27
- Advisory published
- 2026-07-23
- Advisory updated
- 2026-07-27
Who should care
Users of h2oGPT version 0.2.1 and prior should be aware of this critical vulnerability. The vulnerability allows for remote code execution, which can lead to severe consequences, including data breaches and system compromise. Organizations using h2oGPT should prioritize patching this vulnerability to prevent potential attacks. Operators, platform administrators, vulnerability management teams, and security teams should review exposure and implement mitigations.
Technical summary
The vulnerability is caused by the unsanitized use of the bearer token as a path component in the get_user_dir function of openai_server/backend_utils.py. This allows attackers to supply traversal sequences and access arbitrary files on the server. The vulnerability has a CVSS score of 9.3 and is classified as CRITICAL. Affected product deployments of h2oGPT version 0.2.1 and prior should be reviewed for exposure. Users should prioritize patching this vulnerability to prevent potential attacks, including remote code execution by writing to startup hooks or application-loaded files. The CVE record was published on 2026-07-23T18:17:01.327Z and last modified on 2026-07-27T17:16:40.310Z. The NVD entry is currently Deferred.
Defensive priority
High
Recommended defensive actions
- Patch h2oGPT to the latest version
- Restrict access to the OpenAI-compatible files API
- Monitor for suspicious activity
- Implement additional security measures, such as authentication and authorization
- Review compensating controls for exposed systems while remediation is scheduled and verified
- Check relevant monitoring, detection, and logs for exposed assets that need extra review
- Track exceptions, retest remediated assets, and close the item only after evidence is documented
Evidence notes
The CVE record was published on 2026-07-23T18:17:01.327Z and last modified on 2026-07-27T17:16:40.310Z. The NVD entry is currently Deferred. Evidence is limited to CVE and NVD details. Defenders should verify affected product scope, vendor guidance, and potential exposure.
Official resources
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-07-23T18:17:01.327Z and has not been modified since then. The NVD entry is currently Deferred.