CVE-2026-33764 is a medium-severity authorization bypass vulnerability identified in the open-source video platform WWBN AVideo, specifically affecting versions up to and including 26.0. The flaw resides in the AI plugin's save.json.php endpoint, which loads AI response objects based on a user-controlled 'id' parameter from the HTTP request. The vulnerability arises because the endpoint fails to verify that the requested AI response actually belongs to the video specified by the user, allowing an authenticated user with AI permissions to reference arbitrary AI response IDs. Consequently, an attacker can retrieve AI-generated metadata—such as titles, descriptions, keywords, summaries, or full transcriptions—associated with other users' private videos. This constitutes an unauthorized disclosure of potentially sensitive information. The vulnerability is classified under CWE-639 (Authorization Bypass Through User-Controlled Key). Exploitation requires the attacker to be authenticated and possess AI permissions but does not require additional user interaction. The vulnerability does not affect the integrity or availability of the system but compromises confidentiality. The issue has been addressed in a patch (commit aa2c46a806960a0006105df47765913394eec142), though no known exploits are reported in the wild as of now. The CVSS v3.1 base score is 4.3, reflecting a medium severity level due to the limited scope and required privileges.

The primary impact of this vulnerability is the unauthorized disclosure of AI-generated metadata from private videos, which may include sensitive or proprietary information embedded in titles, descriptions, keywords, summaries, or full transcriptions. Organizations relying on WWBN AVideo to host confidential or private video content risk exposure of this data to unauthorized users with AI permissions. While the vulnerability does not allow modification or deletion of content, the leakage of metadata can lead to privacy violations, intellectual property theft, or competitive disadvantage. In regulated industries, such data exposure could also result in compliance violations and reputational damage. Since exploitation requires authentication with AI permissions, the risk is somewhat mitigated by internal access controls; however, insider threats or compromised accounts could leverage this flaw. The absence of known exploits in the wild reduces immediate risk but does not eliminate the potential for future attacks, especially as the patch is not yet universally applied.

To mitigate this vulnerability, organizations should promptly apply the official patch referenced by commit aa2c46a806960a0006105df47765913394eec142 that enforces proper validation of AI response ownership. Until patching is complete, administrators should audit and restrict AI permissions to the minimum necessary users, ideally limiting access to trusted personnel only. Implement monitoring and alerting on unusual access patterns to AI response endpoints to detect potential exploitation attempts. Review and enhance authentication mechanisms to prevent unauthorized access to accounts with AI permissions, including enforcing strong password policies and multi-factor authentication. Additionally, consider isolating sensitive video content or using separate instances for highly confidential data to reduce exposure. Regularly update and review access control policies to ensure that only authorized users can access AI plugin functionalities. Finally, maintain an incident response plan to address potential data leakage incidents stemming from this vulnerability.