{"api_version":"1","generated_at":"2026-05-13T08:26:28+00:00","cve":"CVE-2026-31218","urls":{"html":"https://cve.report/CVE-2026-31218","api":"https://cve.report/api/cve/CVE-2026-31218.json","docs":"https://cve.report/api","cve_org":"https://www.cve.org/CVERecord?id=CVE-2026-31218","nvd":"https://nvd.nist.gov/vuln/detail/CVE-2026-31218"},"summary":{"title":"CVE-2026-31218","description":"The _load_model() function in the neural_magic_training.py script of the optimate project in commit a6d302f912b481c94370811af6b11402f51d377f (2024-07-21) is vulnerable to insecure deserialization (CWE-502). When loading a model state dictionary from a state_dict.pt file via torch.load(), the function does not enable the weights_only=True security parameter. This allows the deserialization of arbitrary Python objects through the Pickle module. A remote attacker can exploit this by providing a maliciously crafted state_dict.pt file within a directory specified via the --model argument, leading to arbitrary code execution during the deserialization process on the victim's system.","state":"PUBLISHED","assigner":"mitre","published_at":"2026-05-12 16:16:13","updated_at":"2026-05-12 16:16:13"},"problem_types":["n/a"],"metrics":[],"references":[{"url":"https://github.com/nebuly-ai/optimate","name":"https://github.com/nebuly-ai/optimate","refsource":"cve@mitre.org","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://www.notion.so/CVE-2026-31218-35d1e139318881839bc8cf6007be2c76","name":"https://www.notion.so/CVE-2026-31218-35d1e139318881839bc8cf6007be2c76","refsource":"cve@mitre.org","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://www.cve.org/CVERecord?id=CVE-2026-31218","name":"CVE Program record","refsource":"CVE.ORG","tags":["canonical"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-31218","name":"NVD vulnerability detail","refsource":"NVD","tags":["canonical","analysis"]}],"affected":[{"source":"CNA","vendor":"n/a","product":"n/a","version":"affected n/a","platforms":[]}],"timeline":[],"solutions":[],"workarounds":[],"exploits":[],"credits":[],"nvd_cpes":[],"vendor_comments":[],"enrichments":{"kev":null,"epss":null,"legacy_qids":[]},"source_records":{"cve_program":{"containers":{"cna":{"affected":[{"product":"n/a","vendor":"n/a","versions":[{"status":"affected","version":"n/a"}]}],"descriptions":[{"lang":"en","value":"The _load_model() function in the neural_magic_training.py script of the optimate project in commit a6d302f912b481c94370811af6b11402f51d377f (2024-07-21) is vulnerable to insecure deserialization (CWE-502). When loading a model state dictionary from a state_dict.pt file via torch.load(), the function does not enable the weights_only=True security parameter. This allows the deserialization of arbitrary Python objects through the Pickle module. A remote attacker can exploit this by providing a maliciously crafted state_dict.pt file within a directory specified via the --model argument, leading to arbitrary code execution during the deserialization process on the victim's system."}],"problemTypes":[{"descriptions":[{"description":"n/a","lang":"en","type":"text"}]}],"providerMetadata":{"dateUpdated":"2026-05-12T15:07:21.158Z","orgId":"8254265b-2729-46b6-b9e3-3dfca2d5bfca","shortName":"mitre"},"references":[{"url":"https://github.com/nebuly-ai/optimate"},{"url":"https://www.notion.so/CVE-2026-31218-35d1e139318881839bc8cf6007be2c76"}]}},"cveMetadata":{"assignerOrgId":"8254265b-2729-46b6-b9e3-3dfca2d5bfca","assignerShortName":"mitre","cveId":"CVE-2026-31218","datePublished":"2026-05-12T00:00:00.000Z","dateReserved":"2026-03-09T00:00:00.000Z","dateUpdated":"2026-05-12T15:07:21.158Z","state":"PUBLISHED"},"dataType":"CVE_RECORD","dataVersion":"5.2"},"nvd":{"publishedDate":"2026-05-12 16:16:13","lastModifiedDate":"2026-05-12 16:16:13","problem_types":["n/a"],"metrics":[],"configurations":[]},"legacy_mitre":{"record":{"CveYear":"2026","CveId":"31218","Ordinal":"1","Title":"CVE-2026-31218","CVE":"CVE-2026-31218","Year":"2026"},"notes":[{"CveYear":"2026","CveId":"31218","Ordinal":"1","NoteData":"The _load_model() function in the neural_magic_training.py script of the optimate project in commit a6d302f912b481c94370811af6b11402f51d377f (2024-07-21) is vulnerable to insecure deserialization (CWE-502). When loading a model state dictionary from a state_dict.pt file via torch.load(), the function does not enable the weights_only=True security parameter. This allows the deserialization of arbitrary Python objects through the Pickle module. A remote attacker can exploit this by providing a maliciously crafted state_dict.pt file within a directory specified via the --model argument, leading to arbitrary code execution during the deserialization process on the victim's system.","Type":"Description","Title":"CVE-2026-31218"}]}}}