Finding and attribution
The ML-DSA signing and verification code copies an unchecked context string into a fixed stack buffer. Anthropic attributes the finding to Claude Mythos Preview; the linked advisory records the affected libgcrypt versions and fix.
Product: libgcrypt. AI system: Claude Mythos Preview. Team / organization: Anthropic / Calif.
CVE identifiers: CVE-2026-41990.
AI and human contribution (source account): Individual Anthropic CVD page explicitly names Claude Mythos Preview.
Human involvement is only described where the source account establishes it; unknown involvement does not mean none. Vendor confirmation of a flaw does not confirm every claim about AI contribution.
Discovery context: Public vulnerability and AI-discovery attribution are documented. Sources do not establish exploitation in the wild or zero-day status.
Remediation status: fix documented. Fixed in libgcrypt 1.12.2.
A source-documented fix does not establish installed patches or current safety. Source checking does not independently reproduce the finding or establish observed exploitation.
Disclosed 23 Apr 2026. AI attribution dated 17 Aug 2026. Sources checked 14 Nov 2026; these are separate dates, not a live installation check.
Recorded CVSS: Medium · highest recorded score 4. Documented technical effects: other.
Severity, consequences and exploitation evidence
Case grouping uses the highest recorded CVSS score, with its version-specific severity band. It is descriptive, not a combined or editorial risk index. All recorded assessments remain below; scores from different versions and sources may disagree. CVSS specifications define version-specific scoring and severity bands.
- CVE-2026-41990: CVSS 3.1 · 4 (Medium) ·
CVSS:3.1/AV:L/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:L· Assessment scenarios: GENERAL. mitre (CNA; CVE Program record) · checked 16 Nov 2026.
CVE-2026-41990: other.
Prerequisites / affected scope: The CNA states that Dilithium signing writes to a static array without a bounds check, but the writes do not use attacker-controlled data. The source does not establish attacker reachability or a further consequence; do not infer attacker-controlled corruption..
Exploitation evidence: The cited source does not establish exploitation in the wild.. Impact source · checked 16 Nov 2026.
Documented fix: Fixed in libgcrypt 1.12.2.