{"id":"CVE-2026-1462","title":"A vulnerability in the `TFSMLayer` class of the `keras` package, version 3.13.0, allows attacker-controlled TensorFlow SavedModels to be loaded during deserialization of `.keras` models, even when `safe_mode=True`","summary":"A vulnerability in the `TFSMLayer` class of the `keras` package, version 3.13.0, allows attacker-controlled TensorFlow SavedModels to be loaded during deserialization of `.keras` models, even when `safe_mode=True`. This bypasses the secu…","severity":"high","cvss":7.8,"cvssVector":"CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","cwe":["CWE-502"],"vendor":"keras","product":"keras","affected":["keras = 3.13.0","openshift_ai >= 2.25, < 2.25.7"],"patched":["openshift_ai 2.25.7"],"published":"2026-04-13","updated":"2026-07-07","source":"NVD","sourceUrl":"https://nvd.nist.gov/vuln/detail/CVE-2026-1462","references":[{"url":"https://github.com/keras-team/keras/commit/b6773d3decaef1b05d8e794458e148cb362f163f","label":"security@huntr.dev"},{"url":"https://huntr.com/bounties/7e78d6f1-6977-4300-b595-e81bdbda331c","label":"security@huntr.dev"},{"url":"https://access.redhat.com/errata/RHSA-2026:24977","label":"0b0ca135-0b70-47e7-9f44-1890c2a1c46c"},{"url":"https://access.redhat.com/security/cve/CVE-2026-1462","label":"0b0ca135-0b70-47e7-9f44-1890c2a1c46c"},{"url":"https://bugzilla.redhat.com/show_bug.cgi?id=2457856","label":"0b0ca135-0b70-47e7-9f44-1890c2a1c46c"},{"url":"https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-1462.json","label":"0b0ca135-0b70-47e7-9f44-1890c2a1c46c"},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-1462"},{"url":"https://github.com/keras-team/keras/pull/22035"},{"url":"https://github.com/keras-team/keras"}],"tags":["nvd","osv","pip"],"epss":0.00405,"epssPercentile":0.3454,"ingestedAt":"2026-07-07T18:42:24.224Z","aliases":["GHSA-4f3f-g24h-fr8m","PYSEC-2026-2547"],"ecosystem":"pip","scores":{"nvd":7.8,"osv":8.8},"slug":"CVE-2026-1462","body":"## Overview\n\nA vulnerability in the `TFSMLayer` class of the `keras` package, version 3.13.0, allows attacker-controlled TensorFlow SavedModels to be loaded during deserialization of `.keras` models, even when `safe_mode=True`. This bypasses the security guarantees of `safe_mode` and enables arbitrary attacker-controlled code execution during model inference under the victim's privileges. The issue arises due to the unconditional loading of external SavedModels, serialization of attacker-controlled file paths, and the lack of validation in the `from_config()` method.\n\n## Affected\n\n- `keras = 3.13.0`\n- `openshift_ai >= 2.25, < 2.25.7`\n\n## Remediation\n\nUpgrade past the affected range:\n\n- `openshift_ai 2.25.7`\n\n## Package advisory (CVE-2026-1462)\n\nAffected packages:\n\n- `keras < 3.13.2`\n\nPatched in:\n\n- `keras 3.13.2`\n\nSource: https://osv.dev/vulnerability/GHSA-4f3f-g24h-fr8m","depth":"twilight","depthScore":43,"depthScoreParts":{"impact":42.9,"likelihood":0.1,"exploitation":0,"ransomware":0},"changes":[]}