# torch (PyPI) vulnerabilities 31 known vulnerabilities (2 critical, 7 high, 18 medium, 3 low), with the versions each affects and the releases that fix them. ## At a glance - Latest release: not known (cve.now hasn't read this package's releases) - Newest unaffected release: not known without release data - Weekly downloads: 14.6M - Dependents: 26.1K - Registry: https://pypi.org/project/torch/ ## Advisories (31) ### [CVE-2026-4538](https://cve.now/CVE-2026-4538): PyTorch pt2 Loading deserialization - Severity: Medium 4.8 - Published: Mar 22, 2026 - Affected: through 2.10.0. ### [CVE-2026-24747](https://cve.now/CVE-2026-24747): PyTorch Vulnerable to Remote Code Execution via Untrusted Checkpoint Files - Severity: High 8.8 - Published: Jan 27, 2026 - Affected: fixed in 2.10.0. ### [CVE-2025-63396](https://cve.now/CVE-2025-63396): An issue was discovered in PyTorch v2.5 and v2.7.1. - Severity: Low 3.3 - Published: Nov 12, 2025 - Affected: through 2.5.0-NA; through 2.7.1-NA. ### [CVE-2025-55560](https://cve.now/CVE-2025-55560): 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… - Severity: High 7.5 - Published: Sep 25, 2025 - Affected: fixed in 2.7.1. ### [CVE-2025-55558](https://cve.now/CVE-2025-55558): A buffer overflow occurs in pytorch v2.7.0 when a PyTorch model consists of torch.nn.Conv2d, torch.nn.functional.hardshrink, and… - Severity: High 7.5 - Published: Sep 25, 2025 - Affected: fixed in 2.7.1. ### [CVE-2025-55557](https://cve.now/CVE-2025-55557): 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… - Severity: High 7.5 - Published: Sep 25, 2025 - Affected: fixed in 2.7.1. ### [CVE-2025-55554](https://cve.now/CVE-2025-55554): pytorch v2.8.0 was discovered to contain an integer overflow in the component torch.nan\_to\_num-.long(). - Severity: Medium 5.3 - Published: Sep 25, 2025 - Affected: fixed in 2.9.0. ### [CVE-2025-55553](https://cve.now/CVE-2025-55553): A syntax error in the component proxy\_tensor.py of pytorch v2.7.0 allows attackers to cause a Denial of Service (DoS). - Severity: High 7.5 - Published: Sep 25, 2025 - Affected: fixed in 2.7.1. ### [CVE-2025-55552](https://cve.now/CVE-2025-55552): pytorch v2.8.0 was discovered to display unexpected behavior when the components torch.rot90 and torch.randn\_like are used together. - Severity: Medium 5.3 - Published: Sep 25, 2025 - Affected: fixed in 2.9.0. ### [CVE-2025-55551](https://cve.now/CVE-2025-55551): 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… - Severity: High 7.5 - Published: Sep 25, 2025 - Affected: fixed in 2.9.0. ### [CVE-2025-46153](https://cve.now/CVE-2025-46153): PyTorch before 3.7.0 has a bernoulli\_p decompose function in decompositions.py even though it lacks full consistency with the eager CPU… - Severity: Medium 5.3 - Published: Sep 25, 2025 - Affected: from 2.6.0, fixed in 2.7.0. ### [CVE-2025-46152](https://cve.now/CVE-2025-46152): In PyTorch before 2.7.0, bitwise\_right\_shift produces incorrect output for certain out-of-bounds values of the "other" argument. - Severity: Medium 5.3 - Published: Sep 25, 2025 - Affected: from 2.6.0, fixed in 2.7.0. ### [CVE-2025-46150](https://cve.now/CVE-2025-46150): In PyTorch before 2.7.0, when torch.compile is used, FractionalMaxPool2d has inconsistent results. - Severity: Medium 5.3 - Published: Sep 25, 2025 - Affected: from 2.6.0, fixed in 2.7.0. ### [CVE-2025-46149](https://cve.now/CVE-2025-46149): In PyTorch before 2.7.0, when inductor is used, nn.Fold has an assertion error. - Severity: Medium 5.3 - Published: Sep 25, 2025 - Affected: from 2.6.0, fixed in 2.7.0. ### [CVE-2025-46148](https://cve.now/CVE-2025-46148): In PyTorch through 2.6.0, when eager is used, nn.PairwiseDistance(p=2) produces incorrect results. - Severity: Medium 5.3 - Published: Sep 25, 2025 - Affected: fixed in 2.7.0. ### [CVE-2025-32434](https://cve.now/CVE-2025-32434): PyTorch: \`torch.load\` with \`weights\_only=True\` leads to remote code execution - Severity: Critical 9.3 - Published: Apr 18, 2025 - Affected: fixed in 2.6.0. ### [CVE-2025-3730](https://cve.now/CVE-2025-3730): PyTorch LossCTC.cpp torch.nn.functional.ctc\_loss denial of service - Severity: Medium 4.8 - Published: Apr 16, 2025 - Affected: fixed in 2.8.0. ### [CVE-2025-3136](https://cve.now/CVE-2025-3136): PyTorch CUDACachingAllocator.cpp torch.cuda.memory.caching\_allocator\_delete memory corruption - Severity: Medium 4.8 - Published: Apr 3, 2025 - Affected: through 2.6.0-NA. ### [CVE-2025-3121](https://cve.now/CVE-2025-3121): PyTorch torch.jit.jit\_module\_from\_flatbuffer memory corruption - Severity: Medium 4.8 - Published: Apr 2, 2025 - Affected: through 2.6.0-NA. ### [CVE-2025-3001](https://cve.now/CVE-2025-3001): PyTorch torch.lstm\_cell memory corruption - Severity: Medium 4.8 - Published: Mar 31, 2025 - Affected: fixed in 2.10.0; through 2.6.0-NA. ### [CVE-2025-3000](https://cve.now/CVE-2025-3000): PyTorch torch.jit.script memory corruption - Severity: Medium 4.8 - Published: Mar 31, 2025 - Affected: fixed in 2.13.0; through 2.6.0-NA. ### [CVE-2025-2999](https://cve.now/CVE-2025-2999): PyTorch torch.nn.utils.rnn.unpack\_sequence memory corruption - Severity: Medium 4.8 - Published: Mar 31, 2025 - Affected: fixed in 2.9.1; through 2.6.0-NA. ### [CVE-2025-2998](https://cve.now/CVE-2025-2998): PyTorch torch.nn.utils.rnn.pad\_packed\_sequence memory corruption - Severity: Medium 4.8 - Published: Mar 31, 2025 - Affected: through 2.6.0; through 2.6.0-NA. ### [CVE-2025-2953](https://cve.now/CVE-2025-2953): PyTorch torch.mkldnn\_max\_pool2d denial of service - Severity: Medium 4.8 - Published: Mar 30, 2025 - Affected: through 2.6.0+cu124; fixed in 2.7.1-rc1. ### [CVE-2025-2149](https://cve.now/CVE-2025-2149): PyTorch Quantized Sigmoid Module nnq\_Sigmoid initialization - Severity: Low 2.0 - Published: Mar 10, 2025 - Affected: through 2.6.0; through 2.6.0-cu124. ### [CVE-2025-2148](https://cve.now/CVE-2025-2148): PyTorch Tuple torch.ops.profiler.\_call\_end\_callbacks\_on\_jit\_fut memory corruption - Severity: Low 2.3 - Published: Mar 10, 2025 - Affected: through 2.6.0-cu124; through 2.6.0. ### [CVE-2024-48063](https://cve.now/CVE-2024-48063): In PyTorch \<=2.4.1, the RemoteModule has Deserialization RCE. - Severity: Unrated - Published: Oct 29, 2024 - Affected: fixed in 2.5.0. ### [CVE-2024-31584](https://cve.now/CVE-2024-31584): Pytorch before v2.2.0 has an Out-of-bounds Read vulnerability via the component torch/csrc/jit/mobile/flatbuffer\_loader.cpp. - Severity: Medium 5.5 - Published: Apr 19, 2024 - Affected: fixed in 2.2.0. ### [CVE-2024-31583](https://cve.now/CVE-2024-31583): Pytorch before version v2.2.0 was discovered to contain a use-after-free vulnerability in torch/csrc/jit/mobile/interpreter.cpp. - Severity: High 7.8 - Published: Apr 17, 2024 - Affected: fixed in 2.2.0; fixed in 2.2.0. ### [CVE-2024-31580](https://cve.now/CVE-2024-31580): PyTorch before v2.2.0 was discovered to contain a heap buffer overflow vulnerability in the component /runtime/vararg\_functions.cpp. - Severity: Medium 4.0 - Published: Apr 17, 2024 - Affected: fixed in 2.2.0; fixed in 2.2.0. ### [CVE-2022-45907](https://cve.now/CVE-2022-45907): In PyTorch before trunk/89695, torch.jit.annotations.parse\_type\_line can cause arbitrary code execution because eval is used unsafely. - Severity: Critical 9.8 - Published: Nov 26, 2022 - Affected: fixed in 1.13.1; fixed in 1.13.1. --- Source: cve.now Built by [Sazabi](https://www.sazabi.com/?utm_source=cve-now&utm_medium=markdown), AI-native observability and security: automatic issue detection, plain-language log search and live dashboards. Page: https://cve.now/packages/pypi/torch Site guide for agents: https://cve.now/llms.txt Uses NVD data; not endorsed by NVD.