CVE-2026-4372

CVE-2026-4372 affects HuggingFace transformers prior to 5.3.0. A malicious config.json can set _attn_implementation_internal to an attacker-controlled HuggingFace Hub repo ID. When a victim loads a model with AutoModelForCausalLM.from_pretrained(), the library downloads and executes arbitrary Pyt...

[CWE-73] Keras — Arbitrary HDF5 File Read via Virtual Dataset Bypass of External Storage Check (CVE-2026-1669 Incomplete Fix)

Description The fix for CVE-2026-1669 in Keras's H5IOStore.verifydataset method savinglib.py:1051 is incomplete. The patch checks dataset.external to block HDF5 datasets that use external file storage, but it does not check dataset.isvirtual. HDF5 Virtual Datasets VDS — created via...

CVE-2026-28500

Open Neural Network Exchange ONNX is an open standard for machine learning interoperability. In versions up to and including 1.20.1, a security control bypass exists in onnx.hub.load due to improper logic in the repository trust verification mechanism. While the function is designed to warn users...

CVE-2026-22807 vLLM affected by RCE via auto_map dynamic module loading during model initialization

vLLM is an inference and serving engine for large language models LLMs. Starting in version 0.10.1 and prior to version 0.14.0, vLLM loads Hugging Face automap dynamic modules during model resolution without gating on trustremotecode, allowing attacker-controlled Python code in a model repo/path ...

CVE-2026-22807

vLLM is an inference and serving engine for large language models LLMs. Starting in version 0.10.1 and prior to version 0.14.0, vLLM loads Hugging Face "automap" dynamic modules during model resolution without gating on "trustremotecode", allowing attacker-controlled Python code in a model...

CVE-2026-22807 vLLM affected by RCE via auto_map dynamic module loading during model initialization

vLLM is an inference and serving engine for large language models LLMs. Starting in version 0.10.1 and prior to version 0.14.0, vLLM loads Hugging Face automap dynamic modules during model resolution without gating on trustremotecode, allowing attacker-controlled Python code in a model repo/path ...

Vulnerability in Keras Model.load_model Leading to Arbitrary Local File Loading and SSRF

...

CVE-2025-12058

The Keras.Model.loadmodel method, including when executed with the intended security mitigation safemode=True, is vulnerable to arbitrary local file loading and Server-Side Request Forgery SSRF. This vulnerability stems from the way the StringLookup layer is handled during model loading from a...

GHSA-36RR-WW3J-VRJV The Keras `Model.load_model` method **silently** ignores `safe_mode=True` and allows arbitrary code execution when a `.h5`/`.hdf5` file is loaded.

Note: This report has already been discussed with the Google OSS VRP team, who recommended that I reach out directly to the Keras team. I’ve chosen to do so privately rather than opening a public issue, due to the potential security implications. I also attempted to use the email address listed i...

GHSA-77WQ-646F-JRM2 Duplicate Advisory: The Keras `Model.load_model` method **silently** ignores `safe_mode=True` and allows arbitrary code execution when a `.h5`/`.hdf5` file is loaded.

Duplicate Advisory This advisory has been withdrawn because it is a duplicate of GHSA-36rr-ww3j-vrjv. This link is maintained to preserve external references. Original Description The Keras Model.loadmodel method can be exploited to achieve arbitrary code execution, even with safemode=True. One c...

PYSEC-2025-76

The Keras Model.loadmodel method can be exploited to achieve arbitrary code execution, even with safemode=True. One can create a specially crafted .keras model archive that, when loaded via Model.loadmodel, will trigger arbitrary code to be executed. This is achieved by crafting a special...

DEBIAN-CVE-2025-9905

The Keras Model.loadmodel method can be exploited to achieve arbitrary code execution, even with safemode=True. One can create a specially crafted .h5/.hdf5 model archive that, when loaded via Model.loadmodel, will trigger arbitrary code to be executed. This is achieved by crafting a special...

CVE-2025-9905

The Keras Model.loadmodel method can be exploited to achieve arbitrary code execution, even with safemode=True. One can create a specially crafted .h5/.hdf5 model archive that, when loaded via Model.loadmodel, will trigger arbitrary code to be executed. This is achieved by crafting a special...

UBUNTU-CVE-2025-9906

The Keras Model.loadmodel method can be exploited to achieve arbitrary code execution, even with safemode=True. One can create a specially crafted .keras model archive that, when loaded via Model.loadmodel, will trigger arbitrary code to be executed. This is achieved by crafting a special...

Deserialization of Untrusted Data

Overview keras is a Keras is a high-level neural networks API for Python.. Affected versions of this package are vulnerable to Deserialization of Untrusted Data via the Model.loadmodel method. An attacker can execute arbitrary code by providing a specially crafted .keras model archive containing ...

CVE-2025-9906 Arbitrary Code execution in Keras Safe Mode

The Keras Model.loadmodel method can be exploited to achieve arbitrary code execution, even with safemode=True. One can create a specially crafted .keras model archive that, when loaded via Model.loadmodel, will trigger arbitrary code to be executed. This is achieved by crafting a special...

PT-2025-38518

Name of the Vulnerable Software and Affected Versions Keras affected versions not specified Description The Model.load model method is susceptible to arbitrary code execution, even when safe mode is enabled. A specially crafted .keras model archive containing a modified config.json file can trigg...

CVE-2025-8747

A safe mode bypass vulnerability in the Model.loadmodel method in Keras versions 3.0.0 through 3.10.0 allows an attacker to achieve arbitrary code execution by convincing a user to load a specially crafted .keras model archive...

Keras vulnerable to CVE-2025-1550 bypass via reuse of internal functionality

Summary It is possible to bypass the mitigation introduced in response to CVE-2025-1550, when an untrusted Keras v3 model is loaded, even when “safemode” is enabled, by crafting malicious arguments to built-in Keras modules. The vulnerability is exploitable on the default configuration and does n...

Duplicate Advisory: Keras safe mode bypass vulnerability

Duplicate Advisory This advisory has been withdrawn because it is a duplicate of GHSA-c9rc-mg46-23w3. This link is maintained to preserve external references. Original Description A safe mode bypass vulnerability in the Model.loadmodel method in Keras versions 3.0.0 through 3.10.0 allows an...