PyTorchPYSEC-2025-206

pytorch v2.8.0 was discovered to contain an integer overflow in the component torch.nan_to_num-.long().

Medium5.3CVE-2025-55554 · Published Sep 25, 2025 · updated May 20, 2026

Source advisory

Affected versions

Package	Affected	Fixed in
torch PyPI	< 2.9.0	2.9.0

Details and references

pytorch v2.8.0 was discovered to contain an integer overflow in the component torch.nan_to_num-.long().

CVSS 3.1: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L
Severity from: the CVSS score
Also known as: BIT-pytorch-2025-55554, CVE-2025-55554

More PyTorch advisories

Date	Advisory	Severity	Fixed in
Sep 252025	In PyTorch through 2.6.0, when eager is used, nn.PairwiseDistance(p=2) produces incorrect results. CVE-2025-46148Medium5.3fixed in 2.7.0	Medium5.3	2.7.0
Sep 252025	In PyTorch before 2.7.0, when inductor is used, nn.Fold has an assertion error. CVE-2025-46149Medium5.3fixed in 2.7.0	Medium5.3	2.7.0
Sep 252025	In PyTorch before 2.7.0, when torch.compile is used, FractionalMaxPool2d has inconsistent results. CVE-2025-46150Medium5.3fixed in 2.7.0	Medium5.3	2.7.0
Sep 252025	In PyTorch before 2.7.0, bitwise_right_shift produces incorrect output for certain out-of-bounds values of the "other" argument. CVE-2025-46152Medium5.3fixed in 2.7.0	Medium5.3	2.7.0
Sep 252025	PyTorch before 3.7.0 has a bernoulli_p 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 fallback_random=True. CVE-2025-46153Medium5.3fixed in 2.7.0	Medium5.3	2.7.0
Sep 252025	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. CVE-2025-55551High7.5fixed in 2.9.0	High7.5	2.9.0

Critical advisories by email

Wednesdays: the week’s critical and high advisories in the AI and data stack, with the fixed versions. Only in weeks that have some.