PyTorch

A remote code execution vulnerability, CVE-2026-58659, exists in PyTorch Lightning through version 2.6.5, allowing attackers to craft malicious checkpoint files that execute arbitrary code by exploiting a flaw in the `_load_state` function when `LightningModule.load_from_checkpoint` is called.

A critical vulnerability (CVE-2025-71342) exists in picklescan versions prior to 0.0.30, where it fails to detect malicious code embedded in Python pickle files by leveraging `idlelib.run.Executive.runcode` in reduce methods, allowing attackers to conceal and execute arbitrary code during `pickle.load` operations, leading to remote code execution (RCE) and potential supply chain attacks, particularly impacting PyTorch models.

Stanza, an NLP library, is vulnerable to remote code execution (CVE-2026-54499) due to an unsafe fallback mechanism when loading PyTorch model files, allowing an attacker who can place a malicious pretrain or model file to achieve arbitrary code execution on systems processing NLP pipelines, leading to credential theft, backdoors, data exfiltration, and lateral movement.

A remote, authenticated attacker can exploit a vulnerability in vllm and PyTorch to cause a denial-of-service condition or potentially achieve remote code execution.