CVE-2026-23526 is a critical privilege escalation vulnerability affecting the open-source computer vision annotation tool CVAT, specifically versions from 1.0.0 up to but not including 2.55.0. The root cause lies in the improper definition of privileges associated with users assigned the 'staff' status. These users can freely modify their own permissions, including elevating themselves to superuser status and joining the admin group. This escalation grants them unrestricted access to all data and administrative functions within the CVAT instance. The vulnerability is classified under CWE-267, which involves privileges defined with unsafe actions. Exploitation requires the attacker to have staff-level access but no further authentication or user interaction is needed, and the attack can be carried out remotely over the network. The CVSS 4.0 base score is 8.5 (high), reflecting the ease of exploitation and the severe impact on confidentiality and integrity. No known exploits are reported in the wild yet. The vulnerability was publicly disclosed on January 21, 2026, and fixed in CVAT version 2.55.0. As a temporary mitigation, organizations are advised to audit the list of users with staff status and revoke this status from any users who should not have superuser privileges. This vulnerability is particularly critical for organizations that handle sensitive or proprietary annotated datasets in CVAT, as unauthorized access could lead to data leakage, manipulation, or disruption of machine learning workflows.

The primary impact of CVE-2026-23526 on European organizations is the potential unauthorized elevation of privileges within CVAT instances, leading to full administrative control over the annotation platform. This can result in unauthorized access to sensitive annotated datasets, which may include proprietary, personal, or regulated data, thus compromising confidentiality. Integrity is also at risk, as attackers could alter or delete annotations, corrupting data quality and undermining machine learning model training and outcomes. Availability is less directly impacted but could be affected if attackers disrupt or disable the CVAT service. Given CVAT's use in computer vision projects across industries such as automotive, healthcare, and security, exploitation could have downstream effects on product development, research, and compliance with data protection regulations like GDPR. The vulnerability's ease of exploitation without user interaction and over the network increases the risk profile, especially in environments where staff status is broadly assigned or insufficiently monitored. European organizations relying on CVAT for critical AI workflows must consider this vulnerability a significant threat to their data security and operational integrity.

1. Immediate patching: Upgrade all CVAT instances to version 2.55.0 or later, where the vulnerability is fixed. 2. Access control audit: Conduct a thorough review of all users assigned staff status in CVAT and revoke this status from any users who do not require elevated privileges. 3. Principle of least privilege: Restrict staff status assignment strictly to trusted administrators and limit the number of users with such privileges. 4. Monitoring and logging: Enable detailed logging of permission changes and monitor for unusual privilege escalation activities within CVAT. 5. Network segmentation: Isolate CVAT servers within secure network segments to reduce exposure to unauthorized access. 6. Authentication hardening: Integrate CVAT with strong authentication mechanisms (e.g., MFA) to reduce the risk of compromised staff accounts. 7. Incident response readiness: Prepare to respond to potential exploitation by having backups of annotation data and a plan to restore integrity if unauthorized changes are detected. 8. User training: Educate administrators and users about the risks of privilege misuse and the importance of safeguarding staff credentials.