CVE-2025-46150: Medium severity PyTorch PyTorch vulnerability

Published Sep 25, 2025
·
Updated

In PyTorch before 2.7.0, when torch.compile is used, FractionalMaxPool2d has inconsistent results.

Affected Software

9 affected componentsFixes available
PyTorch PyTorch<2.7.0
Microsoft cbl2 pytorch 2.0.0-9
Microsoft azl3 pytorch 2.2.2-7
IBM Concert Software<=1.0.0-3.0.0
linuxfoundation Pytorch Python>=2.6.0<2.7.0
debian/pytorch*
debian/pytorch*
debian/pytorch<2.12.0+dfsg2-1
2.12.0+dfsg2-1
debian/pytorch<=1.13.1+dfsg-4, <=2.6.0+dfsg-7
2.14.1+dfsg-1

Remediation

Recommended actions to resolve this vulnerability, in priority order.

  1. Upgrade

    Upgrade debian/pytorch to a version that resolves this vulnerability.

    Fixed in 2.12.0+dfsg2-1
  2. Upgrade

    Upgrade debian/pytorch to a version that resolves this vulnerability.

    Fixed in 2.14.1+dfsg-1

Event History

Sep 25, 2025
CVE Published
via MITRE·12:00 AM
Data Sourced
via MITRE·12:00 AM
Description
Data Sourced
via OSV·12:00 AM
DescriptionSeverity
Data Sourced
via NVD·03:16 PM
RemedyDescriptionSeverityAffected Software
Data Sourced
via OSV·03:16 PM
Affected Software
Oct 2, 2025
Data Sourced
via Microsoft·01:05 AM
DescriptionSeverityWeaknessAffected Software
Sep 22, 2026
Data Sourced
via IBM·12:00 AM
DescriptionAffected Software

Parent advisories

This vulnerability appears in the following advisories.

Frequently Asked Questions

1

What is the severity of CVE-2025-46150?

The severity of CVE-2025-46150 is considered moderate due to inconsistent results in the FractionalMaxPool2d function.

2

How do I fix CVE-2025-46150?

To fix CVE-2025-46150, upgrade PyTorch to version 2.7.0 or later.

3

What is impacted by CVE-2025-46150?

CVE-2025-46150 specifically impacts PyTorch versions prior to 2.7.0 when using the torch.compile feature.

4

What vulnerabilities are related to CVE-2025-46150?

CVE-2025-46150 is related to issues that arise from inconsistent behavior in CUDA operations in PyTorch.

5

Is CVE-2025-46150 exploitable?

CVE-2025-46150 may lead to incorrect outputs, potentially impacting model performance but does not present a direct security exploit.

Contact

SecAlerts Pty Ltd.
132 Wickham Terrace
Fortitude Valley,
QLD 4006, Australia
info@secalerts.co
By using SecAlerts services, you agree to our services end-user license agreement. This website is safeguarded by reCAPTCHA and governed by the Google Privacy Policy and Terms of Service. All names, logos, and brands of products are owned by their respective owners, and any usage of these names, logos, and brands for identification purposes only does not imply endorsement. If you possess any content that requires removal, please get in touch with us.
© 2026 SecAlerts Pty Ltd.
ABN: 70 645 966 203, ACN: 645 966 203