Pytorch vulnerabilitiesRanked by how likely they are to be used.
Pytorch is made by Linuxfoundation. We track 24 published vulnerabilities in it, across 17 versions. None of them is on the list of flaws known to be exploited today. If you run Pytorch, keep it on a version the vendor still supports.
We keep watching the known-exploited list.
Which Pytorch vulnerabilities are being exploited?
This is the list that matters most: flaws criminals are using now.
None of Pytorch's published vulnerabilities is on the list of flaws known to be exploited today. We keep watching that list.
Which Pytorch flaws are most likely to be used?
Ranked by the latest exploit-likelihood forecast: the chance each flaw is used in an attack in the next 30 days. A high severity score on its own doesn't tell you that. Flaws shared by many other makers' products are left out.
- 2.3%CVE-2025-32434Critical
PyTorch is a Python package that provides tensor computation with strong GPU acceleration and deep neural networks built on a tape-based autograd system.
- 1.3%CVE-2022-45907Critical
In PyTorch before trunk/89695, torch.jit.annotations.parse_type_line can cause arbitrary code execution because eval is used unsafely.
- 0.5%CVE-2025-46152Medium
In PyTorch before 2.7.0, bitwise_right_shift produces incorrect output for certain out-of-bounds values of the "other" argument.
- 0.4%CVE-2025-55558High
A buffer overflow occurs in pytorch v2.7.0 when a PyTorch model consists of torch.nn.Conv2d, torch.nn.functional.hardshrink, and torch.Tensor.view-torch.mv() and is compiled by Inductor, leading to a Denial of...
- 0.4%CVE-2025-55551High
An issue in the component torch.linalg.lu of pytorch v2.8.0 allows attackers to cause a Denial of Service (DoS) when performing a slice operation.
- 0.4%CVE-2025-55552High
pytorch v2.8.0 was discovered to display unexpected behavior when the components torch.rot90 and torch.randn_like are used together.
- 0.4%CVE-2025-55560High
An issue in pytorch v2.7.0 can lead to a Denial of Service (DoS) when a PyTorch model consists of torch.Tensor.to_sparse() and torch.Tensor.to_dense() and is compiled by Inductor.
- 0.4%CVE-2025-55553High
A syntax error in the component proxy_tensor.py of pytorch v2.7.0 allows attackers to cause a Denial of Service (DoS).
- 0.4%CVE-2025-55557High
A Name Error occurs in pytorch v2.7.0 when a PyTorch model consists of torch.cummin and is compiled by Inductor, leading to a Denial of Service (DoS).
- 0.4%CVE-2025-46148Medium
In PyTorch through 2.6.0, when eager is used, nn.PairwiseDistance(p=2) produces incorrect results.
Running Pytorch? We'll keep checking.
Looking flaws up one at a time doesn't scale. Tell us what you run once, and we tell you which flaws matter to you, in what order, and keep watching for new ones.
- Tell us what you runPick your software from a list, or let us work it out from your website.
- We match it to every flawOver 400,000 published flaws, checked against your list.
- Ranked by real-world riskA 0 to 100 score, led by evidence criminals are using it.
- Told as soon as we knowAn alert by email, in the app or by webhook to your own tools.
Free plan, no card needed. Nothing to install.
1 new flaw affects software Acme tracks
PyTorch is a Python package that provides tensor computation with strong GPU acceleration and deep neural networks built on a tape-based autograd system. In version 2.5.1 and prior, a Remote Command Execution (RCE) vulnerability exists in PyTorch when loading a model using torch.load with weights_only=True. This issue has been patched in version 2.6.0.
- Matched to
- Pytorch on Acme's software list
- What to do
- Plan in the vendor’s update
Sent to the alert recipients Acme chose. Also in the app and to Acme's webhook.
We track 24 published vulnerabilities listed under the name Pytorch, across 17 versions.
None of Pytorch's vulnerabilities is on the list of flaws known to be exploited today.
Create a Precursor Intelligence account and track Pytorch. We match it against every published flaw, give each one a risk score from 0 to 100 and keep watching, so you hear about new ones as soon as we know, by email, in the app or by webhook.