CVE-2026-27167

Gradio, in versions 4.16.0 through 6.5.x, running outside Hugging Face Spaces enables mocked OAuth routes when OAuth components are used. Visiting /login/huggingface causes the server to fetch its HF token via hugggingface_hub.get_token() and store it in the visitor’s session cookie, which is sig...

CVE-2026-27167 Gradio: Mocked OAuth Login Exposes Server Credentials and Uses Hardcoded Session Secret

Gradio is an open-source Python package designed for quick prototyping. Starting in version 4.16.0 and prior to version 6.6.0, Gradio applications running outside of Hugging Face Spaces automatically enable "mocked" OAuth routes when OAuth components e.g. gr.LoginButton are used. When a user visi...

CVE-2026-27167 Gradio: Mocked OAuth Login Exposes Server Credentials and Uses Hardcoded Session Secret

Gradio is an open-source Python package designed for quick prototyping. Starting in version 4.16.0 and prior to version 6.6.0, Gradio applications running outside of Hugging Face Spaces automatically enable "mocked" OAuth routes when OAuth components e.g. gr.LoginButton are used. When a user visi...

PT-2026-22414

Name of the Vulnerable Software and Affected Versions Gradio versions prior to 6.6.0 Description Gradio is a Python package for rapid prototyping. A flaw exists in the OAuth flow where the redirect to target function does not properly validate the target url query parameter. This allows redirecti...

SafePickle: Robust and Generic ML Detection of Malicious Pickle-Based ML Models

Model repositories such as Hugging Face increasingly distribute machine learning artifacts serialized with Python's pickle format, exposing users to remote code execution RCE risks during model loading. Recent defenses, such as PickleBall, rely on per-library policy synthesis that requires comple...

Hugging Face Smolagents has a Server-Side Request Forgery issue

A weakness has been identified in huggingface smolagents 1.24.0. Impacted is the function requests.get/requests.post of the component LocalPythonExecutor. Executing a manipulation can lead to server-side request forgery. It is possible to launch the attack remotely. The exploit has been made...

GHSA-JXGV-6J54-WWC7 Hugging Face Smolagents has a Server-Side Request Forgery issue

A weakness has been identified in huggingface smolagents 1.24.0. Impacted is the function requests.get/requests.post of the component LocalPythonExecutor. Executing a manipulation can lead to server-side request forgery. It is possible to launch the attack remotely. The exploit has been made...

NVIDIA Megatron Bridge 代码注入漏洞

NVIDIA Megatron Bridge is a component developed by NVIDIA Corporation in the United States that connects Hugging Face and Megatron-Core. NVIDIA Megatron Bridge has a code injection vulnerability. This vulnerability arises from malicious inputs in the data shuffling tutorial, which may lead to cod...

NVIDIA Megatron Bridge 代码注入漏洞

NVIDIA Megatron Bridge is a component developed by NVIDIA Corporation in the United States that connects Hugging Face and Megatron-Core. NVIDIA Megatron Bridge has a code injection vulnerability. This vulnerability arises from malicious inputs in the data merging process, which may lead to code...

Hugging Face Text Generation Inference vulnerable to Uncontrolled Resource Consumption

A vulnerability in huggingface/text-generation-inference version 3.3.6 allows unauthenticated remote attackers to exploit unbounded external image fetching during input validation in VLM mode. The issue arises when the router scans inputs for Markdown image links and performs a blocking HTTP GET...

GHSA-J7X9-7J54-2V3H Hugging Face Text Generation Inference vulnerable to Uncontrolled Resource Consumption

A vulnerability in huggingface/text-generation-inference version 3.3.6 allows unauthenticated remote attackers to exploit unbounded external image fetching during input validation in VLM mode. The issue arises when the router scans inputs for Markdown image links and performs a blocking HTTP GET...

CVE-2026-0599

CVE-2026-0599 concerns huggingface/text-generation-inference version 3.3.6, where unauthenticated attackers can trigger a resource-exhaustion DoS via unbounded external image fetching during input validation in VLM mode. The router scans inputs for Markdown image links and issues a blocking HTTP ...

Securing the AI Supply Chain: What Can We Learn from Developer-Reported Security Issues and Solutions of AI Projects?

The rapid growth of Artificial Intelligence AI models and applications has led to an increasingly complex security landscape. Developers of AI projects must contend not only with traditional software supply chain issues but also with novel, AI-specific security threats. However, little is known...

CVE-2025-14931

Hugging Face smolagents Remote Python Executor Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face smolagents. Authentication is not required to exploit this...

CVE-2025-14928

A flaw was found in the Hugging Face Transformers library. The convertconfig function fails to validate a user-supplied string before using it to execute Python code. An attacker can exploit this flaw by providing a malicious HuBERT model checkpoint, causing arbitrary code execution in the contex...

CVE-2025-14927

A flaw was found in the Hugging Face Transformers library. The convertconfig function fails to validate a user-supplied string before using it to execute Python code. An attacker can exploit this flaw by providing a malicious SEW-D model checkpoint, causing arbitrary code execution in the context...

CVE-2025-14924

A flaw was found in the Hugging Face Transformers library. The parsing of checkpoints fails to validate user-supplied data, causing a deserialization of untrusted data. An attacker can exploit this issue by providing a malicious megatrongpt2 model, resulting in arbitrary code execution in the...

CVE-2025-14920

A flaw was found in the Hugging Face Transformers library. The parsing of model files fails to validate user-supplied data, causing a deserialization of untrusted data. An attacker can exploit this issue by providing a malicious Perceiver model or convincing a user to visit a malicious page,...

CVE-2025-14921

A flaw was found in the Hugging Face Transformers library. The parsing of model files fails to validate user-supplied data, causing a deserialization of untrusted data. An attacker can exploit this issue by providing a malicious Transformer-XL model, resulting in arbitrary code execution in the...

CVE-2025-14922

A flaw was found in Hugging Face Diffusers. This vulnerability allows remote attackers to execute arbitrary code on affected installations. Exploitation requires user interaction, where a target must visit a malicious page or open a malicious file. The issue stems from improper validation of...