CVE-2026-58659
Lightningai Pytorch Lightning ≤ 2.6.5
Raw vector
CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:A/VC:H/VI:H/VA:H/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:XCVSS and EPSS are reproduced from their sources (NVD, FIRST EPSS). Risk Priority is our own derived reading, not an NVD score.
Summary
CVE-2026-58659 is a high-severity Unsafe Reflection (CWE-470) vulnerability in Lightningai Pytorch Lightning. Its CVSS base score is 8.4 (High).
Operationally, exploitation aligns with the MITRE ATT&CK technique Reflective Code Loading (T1620); ranked at the 48th percentile by exploit likelihood (below the median); it is not currently listed in the CISA KEV catalog; a public proof-of-concept is referenced.
This vulnerability is AI-related — categorised as Deep Learning Frameworks; in the Supply Chain and Deployment risk domain.
The strongest mitigations our analysis identified map to SI-10 (Information Input Validation) and AC-3 (Access Enforcement) — see the control section below for these in your framework.
OWASP Top 10 for Web (2025)
EU & UK References
- 🇪🇺 ENISA EUVD: EUVD-2026-44751
Vulnerability Data
PyTorch Lightning through 2.6.5, fixed in commit d710d68, contains a remote code execution vulnerability in the _load_state function that imports and executes attacker-controlled module names from checkpoint _instantiator hyperparameters. Attackers can craft malicious checkpoint files that bypass weights_only=True protections to…
more
execute arbitrary code when LightningModule.load_from_checkpoint is called.
- CWE(s)
AI Security AnalysisAI
- AI Category
- Deep Learning Frameworks
- Risk Domain
- Supply Chain and Deployment
- OWASP Top 10 for LLMs 2025
- None mapped
- Classification Reason
- Matched keywords: pytorch
Related Threats
MITRE ATT&CK Enterprise Techniques
CVEs Like This One
Affected Assets
Mitigating Controls
Mitigating Controls (NIST 800-53 r5) AI
Input validation directly stops externally supplied class or method names from selecting improper code via reflection.
Enforces authorization checks on the code or classes ultimately invoked, blocking unauthorized selections even if reflection is used.
Limits privileges of any code reached through unsafe reflection, reducing blast radius without stopping the selection itself.
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.
Vulnerability identification processes can discover unsafe reflection during code review or scanning.
Preventing execution of unauthorized code can block exploitation of unsafe reflection at runtime.
Secure SDLC practices directly avoid introducing externally controlled class selection via reflection.
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.
A.8.25 mandates secure SDLC practices including secure coding guidelines, security requirements, checkpoints, testing (scans/penetration), and developer training that can catch/prevent unsafe reflection flaws during development; this stops exploitation for a minority slice (first-party code written under the process) but leaves the bulk of the weakness (e.g. in third-party libraries, legacy code, or incomplete enforcement) untouched.
Application security requirements can explicitly prohibit or constrain reflection based on untrusted input.
A.8.27 mandates security-by-design principles (e.g. distrust external input, least privilege, secure session control, data validation/sanitization, design reviews against attack patterns) that reach a slice of unsafe reflection cases but leave a bounded remainder (e.g. reflection in third-party libraries or dynamic class loading where principles are applied incompletely).
Secure coding standards directly forbid unsafe reflection and require whitelisting or static alternatives.