CVE-2026-1462 Safe Mode Bypass in keras-team/keras
- Published
- Updated
- Exploit status
- No public exploit known
- CISA SSVC
- Proof of concept public · automatable: no · technical impact: total
- Technology
- DebianRed Hat
Description
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 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.
Upgrade path
11 packages: 2 fixed, 1 partially fixed, 7 with no fix yet.
| Package | Affected | Fixed in | Status |
|---|---|---|---|
| debian/keras(Debian 11) | = 2.3.1+dfsg-3; = 2.3.1+dfsg2-1; All versions | No fix listed | No fix yet |
| redhat/rhoai/odh-modelmesh-runtime-adapter-rhel8(Red Hat openshift_ai) | All versions | No fix listed | No fix yet |
| redhat/rhoai/odh-modelmesh-runtime-adapter-rhel9:1780394782(Red Hat openshift_ai 2.25 el9) | All versions | No fix listed | No fix yet |
| redhat/rhoai/odh-pipeline-runtime-tensorflow-cuda-py312-rhel9:1782471734(Red Hat openshift_ai 3.3 el9) | All versions | No fix listed | No fix yet |
| redhat/rhoai/odh-pipeline-runtime-tensorflow-rocm-py312-rhel9:1782471879(Red Hat openshift_ai 3.3 el9) | All versions | No fix listed | No fix yet |
| redhat/rhoai/odh-workbench-jupyter-tensorflow-cuda-py312-rhel9:1782471730(Red Hat openshift_ai 3.3 el9) | All versions | No fix listed | No fix yet |
| redhat/rhoai/odh-workbench-jupyter-tensorflow-rocm-py312-rhel9:1782471835(Red Hat openshift_ai 3.3 el9) | All versions | No fix listed | No fix yet |
| redhat openshift ai | < 1780394782; < 1782471730; < 1782471734; < 1782471835 +3 more | 2.25.7, 1780394782, 1782471730, 1782471734, 1782471835, 1782471879 | Partially fixed |
| keras-team/keras | < 3.13.2 | 3.13.2 | Fixed |
| pypi keras(PyPI) | < 3.13.2; = 0.2.0; = 0.3.0; = 0.3.1 +116 more | 3.13.2 | Fixed |
| Keras | = 3.13.0 | No fix listed | Versions listed |
Affected products
| Product | Affected versions | Fixed in |
|---|---|---|
| pypi keras(PyPI) | < 3.13.2 | 3.13.2 |
| pypi keras(PyPI) | = 0.2.0 | No fix listed |
| pypi keras(PyPI) | = 0.3.0 | No fix listed |
| pypi keras(PyPI) | = 0.3.1 | No fix listed |
| pypi keras(PyPI) | = 0.3.2 | No fix listed |
| pypi keras(PyPI) | = 0.3.3 | No fix listed |
| pypi keras(PyPI) | = 1.0.0 | No fix listed |
| pypi keras(PyPI) | = 1.0.1 | No fix listed |
| pypi keras(PyPI) | = 1.0.2 | No fix listed |
| pypi keras(PyPI) | = 1.0.3 | No fix listed |
| pypi keras(PyPI) | = 1.0.4 | No fix listed |
| pypi keras(PyPI) | = 1.0.5 | No fix listed |
| pypi keras(PyPI) | = 1.0.6 | No fix listed |
| pypi keras(PyPI) | = 1.0.7 | No fix listed |
| pypi keras(PyPI) | = 1.0.8 | No fix listed |
| pypi keras(PyPI) | = 1.1.0 | No fix listed |
| pypi keras(PyPI) | = 1.1.1 | No fix listed |
| pypi keras(PyPI) | = 1.1.2 | No fix listed |
| pypi keras(PyPI) | = 1.2.0 | No fix listed |
| pypi keras(PyPI) | = 1.2.1 | No fix listed |
| pypi keras(PyPI) | = 1.2.2 | No fix listed |
| pypi keras(PyPI) | = 2.0.0 | No fix listed |
| pypi keras(PyPI) | = 2.0.1 | No fix listed |
| pypi keras(PyPI) | = 2.0.2 | No fix listed |
| pypi keras(PyPI) | = 2.0.3 | No fix listed |
Show 113 more
| Product | Affected versions | Fixed in |
|---|---|---|
| pypi keras(PyPI) | = 2.0.4 | No fix listed |
| pypi keras(PyPI) | = 2.0.5 | No fix listed |
| pypi keras(PyPI) | = 2.0.6 | No fix listed |
| pypi keras(PyPI) | = 2.0.7 | No fix listed |
| pypi keras(PyPI) | = 2.0.8 | No fix listed |
| pypi keras(PyPI) | = 2.0.9 | No fix listed |
| pypi keras(PyPI) | = 2.1.0 | No fix listed |
| pypi keras(PyPI) | = 2.1.1 | No fix listed |
| pypi keras(PyPI) | = 2.1.2 | No fix listed |
| pypi keras(PyPI) | = 2.1.3 | No fix listed |
| pypi keras(PyPI) | = 2.1.4 | No fix listed |
| pypi keras(PyPI) | = 2.1.5 | No fix listed |
| pypi keras(PyPI) | = 2.1.6 | No fix listed |
| pypi keras(PyPI) | = 2.10.0 | No fix listed |
| pypi keras(PyPI) | = 2.10.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.10.0rc1 | No fix listed |
| pypi keras(PyPI) | = 2.11.0 | No fix listed |
| pypi keras(PyPI) | = 2.11.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.11.0rc1 | No fix listed |
| pypi keras(PyPI) | = 2.11.0rc2 | No fix listed |
| pypi keras(PyPI) | = 2.11.0rc3 | No fix listed |
| pypi keras(PyPI) | = 2.12.0 | No fix listed |
| pypi keras(PyPI) | = 2.12.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.12.0rc1 | No fix listed |
| pypi keras(PyPI) | = 2.13.1 | No fix listed |
| pypi keras(PyPI) | = 2.13.1rc0 | No fix listed |
| pypi keras(PyPI) | = 2.13.1rc1 | No fix listed |
| pypi keras(PyPI) | = 2.14.0 | No fix listed |
| pypi keras(PyPI) | = 2.14.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.15.0 | No fix listed |
| pypi keras(PyPI) | = 2.15.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.15.0rc1 | No fix listed |
| pypi keras(PyPI) | = 2.2.0 | No fix listed |
| pypi keras(PyPI) | = 2.2.1 | No fix listed |
| pypi keras(PyPI) | = 2.2.2 | No fix listed |
| pypi keras(PyPI) | = 2.2.3 | No fix listed |
| pypi keras(PyPI) | = 2.2.4 | No fix listed |
| pypi keras(PyPI) | = 2.2.5 | No fix listed |
| pypi keras(PyPI) | = 2.3.0 | No fix listed |
| pypi keras(PyPI) | = 2.3.1 | No fix listed |
| pypi keras(PyPI) | = 2.4.0 | No fix listed |
| pypi keras(PyPI) | = 2.4.1 | No fix listed |
| pypi keras(PyPI) | = 2.4.2 | No fix listed |
| pypi keras(PyPI) | = 2.4.3 | No fix listed |
| pypi keras(PyPI) | = 2.5.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.6.0 | No fix listed |
| pypi keras(PyPI) | = 2.6.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.6.0rc1 | No fix listed |
| pypi keras(PyPI) | = 2.6.0rc2 | No fix listed |
| pypi keras(PyPI) | = 2.6.0rc3 | No fix listed |
| pypi keras(PyPI) | = 2.7.0 | No fix listed |
| pypi keras(PyPI) | = 2.7.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.7.0rc2 | No fix listed |
| pypi keras(PyPI) | = 2.8.0 | No fix listed |
| pypi keras(PyPI) | = 2.8.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.8.0rc1 | No fix listed |
| pypi keras(PyPI) | = 2.9.0 | No fix listed |
| pypi keras(PyPI) | = 2.9.0rc0 | No fix listed |
| pypi keras(PyPI) | = 2.9.0rc1 | No fix listed |
| pypi keras(PyPI) | = 2.9.0rc2 | No fix listed |
| pypi keras(PyPI) | = 3.0.0 | No fix listed |
| pypi keras(PyPI) | = 3.0.1 | No fix listed |
| pypi keras(PyPI) | = 3.0.2 | No fix listed |
| pypi keras(PyPI) | = 3.0.3 | No fix listed |
| pypi keras(PyPI) | = 3.0.4 | No fix listed |
| pypi keras(PyPI) | = 3.0.5 | No fix listed |
| pypi keras(PyPI) | = 3.1.0 | No fix listed |
| pypi keras(PyPI) | = 3.1.1 | No fix listed |
| pypi keras(PyPI) | = 3.10.0 | No fix listed |
| pypi keras(PyPI) | = 3.11.0 | No fix listed |
| pypi keras(PyPI) | = 3.11.1 | No fix listed |
| pypi keras(PyPI) | = 3.11.2 | No fix listed |
| pypi keras(PyPI) | = 3.11.3 | No fix listed |
| pypi keras(PyPI) | = 3.12.0 | No fix listed |
| pypi keras(PyPI) | = 3.12.1 | No fix listed |
| pypi keras(PyPI) | = 3.12.2 | No fix listed |
| pypi keras(PyPI) | = 3.12.3 | No fix listed |
| pypi keras(PyPI) | = 3.12.4 | No fix listed |
| pypi keras(PyPI) | = 3.13.0 | No fix listed |
| pypi keras(PyPI) | = 3.13.1 | No fix listed |
| pypi keras(PyPI) | = 3.2.0 | No fix listed |
| pypi keras(PyPI) | = 3.2.1 | No fix listed |
| pypi keras(PyPI) | = 3.3.0 | No fix listed |
| pypi keras(PyPI) | = 3.3.1 | No fix listed |
| pypi keras(PyPI) | = 3.3.2 | No fix listed |
| pypi keras(PyPI) | = 3.3.3 | No fix listed |
| pypi keras(PyPI) | = 3.4.0 | No fix listed |
| pypi keras(PyPI) | = 3.4.1 | No fix listed |
| pypi keras(PyPI) | = 3.5.0 | No fix listed |
| pypi keras(PyPI) | = 3.6.0 | No fix listed |
| pypi keras(PyPI) | = 3.7.0 | No fix listed |
| pypi keras(PyPI) | = 3.8.0 | No fix listed |
| pypi keras(PyPI) | = 3.9.0 | No fix listed |
| pypi keras(PyPI) | = 3.9.1 | No fix listed |
| pypi keras(PyPI) | = 3.9.2 | No fix listed |
| keras-team/keras | < 3.13.2 | 3.13.2 |
| redhat openshift ai | < 1780394782 | 1780394782 |
| redhat openshift ai | < 1782471734 | 1782471734 |
| redhat openshift ai | < 1782471879 | 1782471879 |
| redhat openshift ai | < 1782471730 | 1782471730 |
| redhat openshift ai | < 1782471835 | 1782471835 |
| redhat openshift ai | All versions | No fix listed |
| debian/keras(Debian:11) | All versions | No fix listed |
| debian/keras(Debian:11) | = 2.3.1+dfsg-3 | No fix listed |
| debian/keras(Debian:11) | = 2.3.1+dfsg2-1 | No fix listed |
| redhat/rhoai/odh-modelmesh-runtime-adapter-rhel9:1780394782(Red Hat:openshift_ai:2.25::el9) | All versions | No fix listed |
| redhat/rhoai/odh-pipeline-runtime-tensorflow-cuda-py312-rhel9:1782471734(Red Hat:openshift_ai:3.3::el9) | All versions | No fix listed |
| redhat/rhoai/odh-pipeline-runtime-tensorflow-rocm-py312-rhel9:1782471879(Red Hat:openshift_ai:3.3::el9) | All versions | No fix listed |
| redhat/rhoai/odh-workbench-jupyter-tensorflow-cuda-py312-rhel9:1782471730(Red Hat:openshift_ai:3.3::el9) | All versions | No fix listed |
| redhat/rhoai/odh-workbench-jupyter-tensorflow-rocm-py312-rhel9:1782471835(Red Hat:openshift_ai:3.3::el9) | All versions | No fix listed |
| redhat/rhoai/odh-modelmesh-runtime-adapter-rhel8(Red Hat:openshift_ai) | All versions | No fix listed |
| Keras | = 3.13.0 | No fix listed |
| redhat openshift ai | >= 2.25, < 2.25.7 | 2.25.7 |
References
- nvd.nist.gov third-party-advisory vendor-advisory
- github.com patch
- github.com patch
- github.com product
- huntr.com exploit third-party-advisory
- pypi.org product
- github.com vendor-advisory
- Red Hat CVE page mitigation third-party-advisory vdb-entry vendor-advisory
- RHBZ#2457856 issue-tracking third-party-advisory
- security.access.redhat.com third-party-advisory
- RHSA-2026:24977 patch third-party-advisory vendor-advisory
- RHSA-2026:37275 patch vendor-advisory
- security-tracker.debian.org vendor-advisory
- www.cve.org