HighVulnerability

CVE-2026-58659: PyTorch Lightning remote code execution via checkpoint _instantiator

Source: NVD(opens in a new tab)
Identifier: CVE-2026-58659
Published: July 15, 2026
Record updated: October 9, 2026

Affected: lightning < 2.6.6
Fixed in: 2.6.6
Known exploitation: Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
EPSS: 0.6%

Summary

CVE-2026-58659 affects PyTorch Lightning through 2.6.5, fixed in commit d710d68. The _load_state function imports and executes attacker-controlled module names taken from checkpoint _instantiator hyperparameters. A crafted checkpoint that bypasses weights_only=True can run arbitrary code when LightningModule.load_from_checkpoint is called.

Mitigation

Fixed in commit d710d68.

Details

Scores and exploitation

CVSS score: 7.8(high)
Patch available: Yes
CVSS vector: CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Attack vector: local
Attack complexity: low
Privileges required: none
User interaction: required
EPSS: 0.6%Estimated probability that this CVE is exploited in the next 30 days, from the FIRST Exploit Prediction Scoring System. Read on October 13, 2026.
Disclosure date: July 15, 2026

Weaknesses and techniques

Weakness (CWE): CWE-470
AI threat technique (MITRE ATLAS): AML.T0010

Affected

Packages: lightning@< 2.6.6 (fixed: 2.6.6)

Classification

Attack type: Supply chain
Attack sophistication: Moderate
AI component targeted: Framework
Impact: Confidentiality,
Integrity,
Availability
Labels: security

How this record was produced

Summaries, mitigations and labels on this page are generated automatically from the source. The linked source is authoritative. The source gives this advisory a description and no title, so the title above is a headline generated from the source text.

Original source: https://nvd.nist.gov/vuln/detail/CVE-2026-58659

First tracked: July 16, 2026 at 02:07 AM ET, from NVD/CVE Database

Classifier: claude-haiku-5-5, prompt v4

Classifier confidence: 95% (self-assessed)

Affected packages in the Exposure Registry

Matched by package name and ecosystem. Each entry shows whether the package delegates to a language model and how many tracked packages depend on it.

lightningPyPILLM dependency since 2025-10-26 · 3 tracked dependents

Summary

Mitigation

Affected packages in the Exposure Registry

Related items