Tools AI Risk Radar ai-incident-0118
Incident record
Claude Code could skip an organisation's managed security settings
- Severity
- Low
- Status
- Patched
- Type
- Infra Vuln
- Target
- Claude Code for Claude Enterprise from 2.0.68 and Claude for Work (Team) from 2.1.38, before 2.1.260
- Actor
- researcher
- CVE
- CVE-2026-103012
What happened
Claude Code chose a locally stored API key ahead of a user's valid Claude Enterprise or Team sign-in when it fetched the organisation's server-managed settings. When the settings endpoint rejected that key, the session started without the organisation's managed policy, such as permission deny rules, model restrictions and managed-only locks, or kept applying a stale cached copy, while still working as the organisation's account. CVE-2026-103012, published by Anthropic on 30 September 2026, scores it CVSS 4.0 2.0, because it needs local access to a device holding such a key.
Claude Code 2.1.260 fixes it, and Anthropic says users on standard auto-update already have the fix. Settings delivered by MDM or by file were unaffected. The record credits Tamas Voros of the NVIDIA AI Red Team. It is one of three CVE records Anthropic published as a CVE numbering authority on 30 September; the other two cover iperf3 flaws credited to Anthropic as finder.
Sources
- NVD record CVE-2026-103012nvd.nist.gov/vuln/detail/CVE-2026-103012
- Claude Code advisory GHSA-gfvf-j8jh-jxxwgithub.com/anthropics/claude-code/security/advisories/GHSA-g…
One record from the AI Risk Radar, maintained by the Security Desk.Data: CSV · JSON ·RSS · CC BY 4.0 with attribution to YFarmX.Tracker updated · 13 October 2026