HighVulnerability
CVE-2026-24747: PyTorch weights_only unpickler memory corruption via malicious checkpoint file
- Source
- NVD(opens in a new tab)
- Identifier
- CVE-2026-24747
- Published
- Record updated
- Affected
- pytorch < 2.10.0
- Fixed in
- 2.10.0
- Known exploitation
- Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
- EPSS
- 0.8%
Summary
CVE-2026-24747 affects PyTorch versions prior to 2.10.0. A flaw in the `weights_only` unpickler lets an attacker craft a malicious checkpoint file (`.pth`) that, when loaded with `torch.load(..., weights_only=True)`, can corrupt memory and potentially lead to arbitrary code execution. The weakness is classified as CWE-94 and CWE-502.
Mitigation
Version 2.10.0 fixes the issue.
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.
- pytorchPyPINo LLM dependency detected
Related items
- HighCVE-2026-123019: eosphoros-ai DB-GPT untrusted code inclusion via trust_remote_codeSimilar attack · NVD/CVE Database
- MediumResearchers Uncover Prompt Adware Hidden in OpenAI Codex Jailbreak ToolkitSimilar attack · The Hacker News
- InfoA Bag of Tricks for Defending Deep Learning-based Face Recognition Pipelines against Backdoor AttacksSimilar attack · Arxiv (cs.RO + cs.CV security)
- InfoTop agent skill security resources — November 2026Similar attack · Adversa AI Blog
- HighCVE-2026-118627: MLflow SHAP load_explainer untrusted deserialization enables code executionSimilar attack · NVD/CVE Database