AI Science Tracker

Rocq guard checker accepts recursive calls through a higher-order fixpoint

Rocq Prover (formerly Coq) · AI involvement attributed to Claude Opus 4.6

Finding and attribution

The guard checker may accept a fixpoint passed as a higher-order argument to another fixpoint, which then applies it to a non-subterm of its structural argument. The issue can create a proof of False (including 0 = 1) through definitional computation without axioms, tactics, plugins or unsafe flags. This undermines proofs checked by affected Rocq/Coq versions. The maintainer issue says discovered autonomously by Opus 4.6 and includes its unedited report.

Product: Rocq Prover (formerly Coq). AI system: Claude Opus 4.6. Team / organization: Anthropic.

CVE identifiers: CVE-2026-72705.

AI and human contribution (source account): The maintainer’s original issue explicitly says “found autonomously by Opus 4.6” and reproduces the model’s unedited report. The later oss-security notice also credits Claude Opus 4.6 and human researcher Tristan Stérin as discoverers during a fuzzing campaign.

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: The cited primary disclosure documents the issue; it does not establish exploitation before disclosure. The classification records the ordinary public disclosure, not a determination that the flaw was not a zero-day.

Remediation status: fix documented. Fixed by PR #21684 and in Rocq 9.2.0 or later; `coqchk` is described as defense in depth. The PR was publicly merged Mar 4, 2026.

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 28 Feb 2026. AI attribution dated 28 Feb 2026. Sources checked 16 Nov 2026; these are separate dates, not a live installation check.

Recorded CVSS: High · highest recorded score 7.1. 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-2026-72705: CVSS 3.1 · 7.1 (High) · CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:C/C:N/I:H/A:N. oss-security disclosure · checked 16 Nov 2026.

CVE-2026-72705: other.
Prerequisites / affected scope: A malicious or erroneous Rocq source or dependency must be compiled/checked by a vulnerable version; the source describes local input/user interaction, not network exploitation..
Exploitation evidence: The source describes proof-checker unsoundness; no exploitation outside proof construction is reported.. Impact source · checked 16 Nov 2026.

Documented fix: Fixed by PR #21684 and in Rocq 9.2.0 or later; `coqchk` is described as defense in depth. The PR was publicly merged Mar 4, 2026.

Original evidence

Download this source record (JSON)

Open this case in the interactive tracker · Browse all static security cases