- ID
- AIA-014
- Category
- AI AI agent incident
- Disclosed
- 2025-08
- Systems
- Cursor
- Harm type
- Code exec / supply chain
- Status
- fixed
Check Point Research says it disclosed the issue on July 16, 2025, that Cursor issued version 1.3 on July 29, 2025, and that "our independent testing confirms that the issue has been effectively addressed" (release notes did not mention it). - CVE
CVE-2025-54136- CISA KEV
- Not in CISA KEV as of catalog 2026.10.30
Checked 2026-10-31 against catalog 2026.10.30
Summary
Check Point Research reported that "once an MCP is approved, future modifications to its command or arguments are trusted without any additional validation or prompt," which could let a changed shared config run commands persistently.
Controls
Controls that could prevent, detect, or support recovery from this kind of failure. Mappings are editorial judgments made by the method, not findings about what any party should have done.
NIST AI RMF
- MAP 4.1 Risks from third-party components are mapped
- MEASURE 2.7 Security and resilience are evaluated and documented
NIST SP 800-53
- CM-3 Configuration Change Control
- CM-5 Access Restrictions for Change
- SI-7 Software, Firmware, and Information Integrity
Human-approval (JCR) checkpoint
partial: MCP configuration approval, bypassed by later changes
Sources
- research.checkpoint.com primary link checked 2026-10-31, HTTP 200
Post dated August 5, 2025
https://research.checkpoint.com/2025/cursor-vulnerability-mcpoison/
Corrections
No corrections recorded.
To report an error in this entry, see corrections.