CISA list

● PyPI package

torch

Torch has 31 security advisories.

As of , torch has 31 security advisories.

Your version, and the version to upgrade to

Find the row your installed version falls in. The second column is the first release after it that none of the advisories lists as affected; a release outside every row is one no advisory lists. That is what the advisories say, not a promise that a release is free of flaws.
If you runUpgrade toAdvisories
Before 2.13.02.13.028 advisories

10 advisories give their versions in a form this page does not read, so the table may leave out some of their ranges.

Security advisories, newest first

Security advisories, newest first
AdvisoryWhat it isAffectsFixed inPublished
CVE-2026-4538PYSEC-2026-139A vulnerability was identified in PyTorch 2.10.0.Up to and including 2.10.0No fixed version listed
CVE-2026-24747GHSA-63cw-57p8-fm3p, PYSEC-2026-1856, PYSEC-2026-2286PyTorch Vulnerable to Remote Code Execution via Untrusted Checkpoint FilesHigh severityBefore 2.10.02.10.0
CVE-2025-63396PYSEC-2025-210An issue was discovered in PyTorch v2.5 and v2.7.1.This page does not read the form in which it gives its versions.No fixed version listed
CVE-2025-55560PYSEC-2025-209An issue in pytorch v2.7.0 can lead to a Denial of Service (DoS) when a PyTorch model consists of torch.Tensor.tosparse() and torch.Tensor.todense() and is compiled by Inductor.Before 2.7.12.7.1
CVE-2025-55558PYSEC-2025-208A 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 Service (DoS).Before 2.7.12.7.1
CVE-2025-55557PYSEC-2025-207A 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).Before 2.7.12.7.1
CVE-2025-55554PYSEC-2025-206pytorch v2.8.0 was discovered to contain an integer overflow in the component torch.nantonum-.long().Before 2.9.02.9.0
CVE-2025-55553PYSEC-2025-205A syntax error in the component proxytensor.py of pytorch v2.7.0 allows attackers to cause a Denial of Service (DoS).Before 2.7.12.7.1
CVE-2025-55552PYSEC-2025-204pytorch v2.8.0 was discovered to display unexpected behavior when the components torch.rot90 and torch.randnlike are used together.Before 2.9.02.9.0
CVE-2025-55551PYSEC-2025-203An 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.Before 2.9.02.9.0
CVE-2025-46153PYSEC-2025-202PyTorch before 3.7.0 has a bernoullip decompose function in decompositions.py even though it lacks full consistency with the eager CPU implementation, negatively affecting nn.Dropout1d, nn.Dropout2d, and nn.Dropout3d for...From 2.6.0 before 2.7.02.7.0
CVE-2025-46152PYSEC-2025-201In PyTorch before 2.7.0, bitwiserightshift produces incorrect output for certain out-of-bounds values of the "other" argument.From 2.6.0 before 2.7.02.7.0
CVE-2025-46150PYSEC-2025-200In PyTorch before 2.7.0, when torch.compile is used, FractionalMaxPool2d has inconsistent results.From 2.6.0 before 2.7.02.7.0
CVE-2025-46149PYSEC-2025-199In PyTorch before 2.7.0, when inductor is used, nn.Fold has an assertion error.From 2.6.0 before 2.7.02.7.0
CVE-2025-46148PYSEC-2025-198In PyTorch through 2.6.0, when eager is used, nn.PairwiseDistance(p=2) produces incorrect results.Before 2.7.02.7.0
CVE-2025-32434GHSA-53q9-r3pm-6pq6, PYSEC-2025-41PyTorch: `torch.load` with `weights_only=True` leads to remote code executionCritical severityBefore 2.6.02.6.0
CVE-2025-3730GHSA-887c-mr87-cxwp, PYSEC-2026-1970PyTorch Improper Resource Shutdown or Release vulnerabilityModerate severityBefore 2.8.02.8.0
CVE-2025-3136PYSEC-2025-197A vulnerability, which was classified as problematic, has been found in PyTorch 2.6.0.This page does not read the form in which it gives its versions.No fixed version listed
CVE-2025-3121PYSEC-2025-196A vulnerability classified as problematic has been found in PyTorch 2.6.0.This page does not read the form in which it gives its versions.No fixed version listed
CVE-2025-3001GHSA-qfhq-4f3w-5fph, PYSEC-2025-195PyTorch is vulnerable to memory corruption through its torch.lstm_cell functionLow severityBefore 2.10.02.10.0
CVE-2025-3000GHSA-rrmf-rvhw-rf47, PYSEC-2025-194PyTorch is vulnerable to memory corruption through its torch.jit.script functionLow severityBefore 2.13.02.13.0
CVE-2025-2999GHSA-vgrw-7cvw-pwgx, PYSEC-2025-193PyTorch is vulnerable to memory corruption through its unpack_sequence functionModerate severityBefore 2.9.12.9.1
CVE-2025-2998GHSA-f4hp-rmr7-r7v8, PYSEC-2025-192PyTorch is Vulnerable to Memory Consumption through pad_packed_sequence FunctionModerate severityUp to and including 2.6.0No fixed version listed
CVE-2025-2953GHSA-3749-ghw9-m3mg, PYSEC-2025-191PyTorch susceptible to local Denial of ServiceLow severityBefore 2.7.1-rc12.7.1-rc1
CVE-2025-2149GHSA-x3gm-94wq-g975, PYSEC-2025-190PyTorch: Manipulation of the argument scale/zero_point leads to improper initialization via Quantized Sigmoid ModuleLow severityUp to and including 2.6.0No fixed version listed
CVE-2025-2148GHSA-c678-jfcj-6jmf, PYSEC-2025-189PyTorch Tuple Handler is Vulnerable to Memory Corruption through Manipulation of None ArgumentLow severityUp to and including 2.6.0No fixed version listed
CVE-2024-48063PYSEC-2024-259In PyTorch <=2.4.1, the RemoteModule has Deserialization RCE.Before 2.5.02.5.0
CVE-2024-31584PYSEC-2024-250Pytorch before v2.2.0 has an Out-of-bounds Read vulnerability via the component torch/csrc/jit/mobile/flatbufferloader.cpp.Before 2.2.02.2.0
CVE-2024-31580GHSA-5pcm-hx3q-hm94, PYSEC-2024-252, PYSEC-2024-328PyTorch heap buffer overflow vulnerabilityHigh severityBefore 2.2.02.2.0
CVE-2024-31583GHSA-pg7h-5qx3-wjr3, PYSEC-2024-251Pytorch use-after-free vulnerabilityHigh severityBefore 2.2.02.2.0
CVE-2022-45907GHSA-47fc-vmwq-366v, PYSEC-2022-43015PyTorch vulnerable to arbitrary code executionCritical severityBefore 1.13.11.13.1

Other PyPI packages

Sources
  • Package advisories from the GitHub Advisory Database (CC BY 4.0). Version ranges are reworded as sentences (source). Licence: CC BY 4.0
  • Package advisories from the PyPA advisory database (CC BY 4.0). Version ranges are reworded as sentences (source). Licence: CC BY 4.0