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...

Agent Security Is a Systems Problem

We take the position that agent security must be approached as a systems problem: the AI model powering the agent must be treated as an untrusted component, and security invariants must be enforced at the system level. Through this lens, efforts to increase model robustness the dominant viewpoint...

Memory Allocation with Excessive Size Value

Overview org.apache.opennlp:opennlp-tools is an is a machine learning based toolkit for the processing of natural language text. Affected versions of this package are vulnerable to Memory Allocation with Excessive Size Value in the AbstractModelReader class. An attacker can cause the application ...

OpenClaw: Incomplete host-env-security-policy allows untrusted model to substitute compiler binaries via env overrides

Summary Incomplete host-env-security-policy.json allows untrusted model to substitute compiler binaries CC, CXX, CARGOBUILDRUSTC, CMAKECCOMPILER via env overrides on approved host exec requests Current Maintainer Triage - Status: narrow - Normalized severity: medium - Assessment: Shipped v2026.3....

CVE-2026-28500 ONNX Untrusted Model Repository Warnings Suppressed by silent=True in onnx.hub.load() — Silent Supply-Chain Attack

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...

EUVD-2026-0750

MessagePack for Java Vulnerable to Remote DoS via Malicious EXT Payload Allocation...

CVE-2026-21452

MessagePack for Java is a serializer implementation for Java. A denial-of-service vulnerability exists in versions prior to 0.9.11 when deserializing .msgpack files containing EXT32 objects with attacker-controlled payload lengths. While MessagePack-Java parses extension headers lazily, it later...

PYSEC-2025-212

Hugging Face Transformers Transformer-XL Model Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this...

CVE-2025-14921 Hugging Face Transformers Transformer-XL Model Deserialization of Untrusted Data Remote Code Execution Vulnerability

Hugging Face Transformers Transformer-XL Model Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this...

(0Day) Hugging Face Transformers Transformer-XL Model Deserialization of Untrusted Data Remote Code Execution Vulnerability

This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The specific flaw exists within the...

EUVD-2023-32213

Malicious code in bioql PyPI...

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...

Privacy-Preserving LLM Interaction with Socratic Chain-Of-Thought Reasoning and Homomorphically Encrypted Vector Databases

Large language models LLMs are increasingly used as personal agents, accessing sensitive user data such as calendars, emails, and medical records. Users currently face a trade-off: They can send private records, many of which are stored in remote databases, to powerful but untrusted LLM providers...

CVE-2021-41127

Rasa is an open source machine learning framework to automate text-and voice-based conversations. In affected versions a vulnerability exists in the functionality that loads a trained model tar.gz file which allows a malicious actor to craft a model.tar.gz file which can overwrite or replace bot...

Remote Code Execution (RCE)

PyTorch is vulnerable to Remote Command Execution RCE. The vulnerability is due to unsafe deserialization due to the use of torch.loadweightsonly=True on untrusted model files, allowing an attacker to execute arbitrary code by supplying a maliciously crafted model...

Deserialization of Untrusted Data

Overview rasa-pro is a State-of-the-art open-core Conversational AI framework for Enterprises that natively leverages generative AI for effortless assistant development. Affected versions of this package are vulnerable to Deserialization of Untrusted Data by loading a maliciously crafted model in...

PYSEC-2024-229

Hugging Face Transformers Trax Model Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this vulnerability in th...

Design/Logic Flaw

A malformed DLC can trigger Memory Corruption in SNPE library due to out of bounds read, such as by loading an untrusted model e.g. from a remote source...

CVE-2023-28543

CVE-2023-28543 affects Qualcomm SNPE library via a malformed digit-length description language (DLC) that can trigger an out-of-bounds read leading to memory corruption when loading an untrusted model (e.g., from remote sources). The vulnerability is described across multiple records (NVD, Red Ha...

CVE-2023-28543 Out of Bounds read in SNPE Library

A malformed DLC can trigger Memory Corruption in SNPE library due to out of bounds read, such as by loading an untrusted model e.g. from a remote source...