PatchSiren cyber security CVE debrief
CVE-2026-58659 Lightning‑AI CVE debrief
CVE-2026-58659 is a remote code execution vulnerability in PyTorch Lightning through 2.6.5. The vulnerability is caused by the _load_state function importing and executing attacker-controlled module names from checkpoint _instantiator hyperparameters. Attackers can craft malicious checkpoint files that bypass weights_only=True protections to execute arbitrary code when LightningModule.load_from_checkpoint is called. This vulnerability has a high CVSS score of 8.4, indicating a high severity level. Users of PyTorch Lightning versions up to 2.6.5 should be aware of this vulnerability and take immediate action to protect themselves.
- Vendor
- Lightning‑AI
- Product
- PyTorch Lightning
- CVSS
- HIGH 8.4
- CISA KEV
- Not listed in stored evidence
- Original CVE published
- 2026-07-15
- Original CVE updated
- 2026-07-18
- Advisory published
- 2026-07-15
- Advisory updated
- 2026-07-18
Who should care
Users of PyTorch Lightning versions up to 2.6.5, operators, platform administrators, vulnerability management teams, and security teams should be aware of this vulnerability and take immediate action to protect themselves. This vulnerability has a high CVSS score of 8.4, indicating a high severity level. Affected deployments should be identified, and owners should be assigned for follow-up.
Technical summary
The vulnerability exists in the _load_state function of PyTorch Lightning, which imports and executes module names from checkpoint _instantiator hyperparameters. An attacker can create malicious checkpoint files that bypass the weights_only=True protection, allowing for arbitrary code execution when LightningModule.load_from_checkpoint is called. This vulnerability has a high CVSS score of 8.4, indicating a high severity level. The affected product is PyTorch Lightning versions up to 2.6.5.
Defensive priority
High priority should be given to updating PyTorch Lightning to a version beyond 2.6.5, as this vulnerability allows for remote code execution. Compensating controls should be implemented to monitor and restrict access to checkpoint files.
Recommended defensive actions
- Update PyTorch Lightning to a version beyond 2.6.5
- Implement compensating controls to monitor and restrict access to checkpoint files
- Conduct regular inventory checks to ensure all instances are updated
- Monitor for suspicious activity related to checkpoint file loading
- Review relevant monitoring, detection, and logs for exposed assets that need extra review
- Track exceptions, retest remediated assets, and close the item only after evidence is documented
- Confirm whether affected product deployments exist in managed environments and assign an owner for follow-up
Evidence notes
The CVE record was published on 2026-07-15T18:16:48.743Z and was last modified on 2026-07-18T02:17:10.147Z. The NVD entry is currently Awaiting Analysis. There is limited information available about the vulnerability, and defenders should verify the affected scope and severity. The CVE record was created based on limited source details, and additional information may be needed to fully understand the vulnerability.
Official resources
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-07-15T18:16:48.743Z and has not been modified since then. The NVD entry is currently Awaiting Analysis.