CVE-2026-23526
Cvat Computer Vision Annotation Tool 1.0.0 – 2.55.0
Raw vector
CVSS:4.0/AV:N/AC:L/AT:N/PR:H/UI:N/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:XSummary
CVE-2026-23526 is a high-severity Privilege Defined With Unsafe Actions (CWE-267) vulnerability in Cvat Computer Vision Annotation Tool. Its CVSS base score is 8.5 (High).
Operationally, exploitation aligns with the MITRE ATT&CK technique Access Token Manipulation (T1134); ranked at the 17th percentile by exploit likelihood (below the median); it is not currently listed in the CISA KEV catalog.
The strongest mitigations our analysis identified map to AC-6 (Least Privilege) and AC-5 (Separation of Duties) — see the control section below for these in your framework.
Deeper analysis AI-assisted summary
Synthesised by an AI model from the NVD description and linked references — a reading aid, not an authoritative source.
CVE-2026-23526 is a privilege escalation vulnerability in CVAT, an open-source interactive video and image annotation tool for computer vision. It affects versions 1.0.0 through 2.54.0, where users with staff status can arbitrarily modify their own permissions, including elevating themselves to superuser status and joining the admin group. This grants full access to all data within the CVAT instance. The flaw is tracked under CWE-267 (Privilege Defined With Unsafe Actions) and carries a CVSS v3.1 base score of 8.8 (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H).
Any authenticated user with staff status can exploit this vulnerability remotely over the network, requiring low privileges, low attack complexity, and no user interaction. Exploitation allows the attacker to achieve high impacts across confidentiality, integrity, and availability, effectively compromising the entire CVAT instance by accessing, modifying, or disrupting all stored data.
CVAT version 2.55.0 addresses the issue with a fix. As a temporary workaround, administrators should review all users with staff status and revoke it from those not intended to hold superuser privileges. Additional details are available in the GitHub security advisory (https://github.com/cvat-ai/cvat/security/advisories/GHSA-7pvv-w55f-qmw7) and the patching commit (https://github.com/cvat-ai/cvat/commit/88ac7aa4d5b52271a30f1aa387c0f5745f8f77d4).
Given CVAT's use in computer vision annotation pipelines, this vulnerability holds relevance for AI/ML environments handling sensitive datasets. No public reports of real-world exploitation were noted at publication on 2026-01-21.
EU & UK References
- 🇪🇺 ENISA EUVD: EUVD-2026-3773
Vulnerability Data
CVAT is an open source interactive video and image annotation tool for computer vision. In versions 1.0.0 through 2.54.0, users that have the staff status may freely change their permissions, including giving themselves superuser status and joining the admin group,…
more
which gives them full access to the data in the CVAT instance. Version 2.55.0 fixes the issue. As a workaround, review the list of users with staff status and revoke it from any users that are not expected to have superuser privileges.
- CWE(s)
Related Threats
MITRE ATT&CK Enterprise Techniques
CVEs Like This One
Affected Assets
Mitigating Controls
Control response
—
—
- 10 hardening rules · 4 OS baselines
V13.2.2V3.7.3V9.2.2V9.2.3
Mitigating Controls (NIST 800-53 r5) AI
Least privilege directly requires that privileges grant only the minimal actions needed, preventing definition of overly broad privileges that include unintended unsafe operations.
Separation of duties splits high-risk actions across multiple privileges so that no single privilege can perform the full unsafe sequence.
Mitigating Controls (NIST CSF 2.0) AI
Derived directly from the weakness types (CWEs) cited in the NVD entry via our AI-authored CWE→CSF cross-walk (authority under review) — links open the control.
Defining and enforcing least-privilege permissions directly reduces the chance that a privilege permits unintended unsafe actions.
Mitigating Controls (ISO/IEC 27001:2022 Annex A) AI
Derived directly from the weakness types (CWEs) cited in the NVD entry via our AI-authored CWE→ISO cross-walk (authority under review) — links open the control.
Use of privileged utility programs restricts dangerous tools but does not address unsafe actions within assigned privileges.
Access control policy can limit unsafe privilege use but does not define safe actions for each privilege.
Access rights assignment can restrict privileges to intended actions, yet does not guarantee the privilege itself is safe.
Segregation of duties reduces risk of unsafe actions by splitting privileges but does not address unsafe privilege definitions.
Privileged access rights control can restrict unsafe privileges, yet does not ensure the privilege definition itself is safe.
Secure development lifecycle can catch unsafe privilege definitions during design but does not prevent them at runtime.
Hardening callouts derived
Configuration rules from DISA STIG baselines that bear on weaknesses of the type cited by this CVE. Each rule is shown with the relationship its mapping actually records, against the CWE it was authored against. Derived via CVE→CWE over `controls_xwalks` (authoritative rows only; rows rated `none` are excluded).
Oracle Linux 9 (1 rule)
- V-271725 OL 9 must require users to provide a password for privilege escalation. prevents CWE-267
Ubuntu 22.04 (3 rules)
- V-260470 Ubuntu 22.04 LTS, when booted, must require authentication upon booting into single-user and maintenance modes. prevents CWE-267
- V-260529 Ubuntu 22.04 LTS must be configured so that remote X connections are disabled, unless to fulfill documented and validated mission requirements. prevents CWE-267
- V-260542 Ubuntu 22.04 LTS must prevent direct login into the root account. prevents CWE-267