CVE-2026-41523

A flaw was found in vLLM, an inference and serving engine for large language models LLMs. An unauthenticated attacker can exploit an assert-based security check during activation function loading. By publishing a malicious HuggingFace model, an attacker can achieve arbitrary code execution on the...

Out Of Band Data Exfiltration

Claude Code is vulnerable to Out-of-Band Data Exfiltration. The vulnerability is due to the pre-approval of the hostname huggingface.co as a bare hostname for the WebFetch tool, where any path on that domain—including attacker-controlled model repositories—was auto-approved without a permission...

CVE-2026-54316

Claude Code exposed an out-of-band data exfiltration risk due to pre-approved huggingface.co as a bare hostname for WebFetch in versions 0.2.54–2.1.163. An attacker who could inject untrusted content into a Claude Code context could force WebFetch to access attacker-controlled model repos (e.g., ...

CVE-2026-54316 Claude Code: Out-of-Band Data Exfiltration via Pre-Approved HuggingFace Domain in WebFetch

Claude Code is an agentic coding tool. From 0.2.54 until 2.1.163, because the hostname huggingface.co was pre-approved as a bare hostname for the WebFetch tool, any path on that domain—including attacker-controlled model repositories—was auto-approved without a permission prompt or being subject ...

CVE-2026-41523

vLLM is an inference and serving engine for large language models LLMs. Prior to 0.22.0, an assert-based security check in vLLM's activation function loading allows any unauthenticated attacker to achieve arbitrary code execution on the server by publishing a malicious HuggingFace model, when vLL...

CVE-2026-41523

vLLM is an inference and serving engine for large language models LLMs. Prior to 0.22.0, an assert-based security check in vLLM's activation function loading allows any unauthenticated attacker to achieve arbitrary code execution on the server by publishing a malicious HuggingFace model, when vLL...

Malicious code in zod-pino (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector c536e5a7ee3d5542e1ac822b30ba4525e52b2ae0c964d0c2470468d91b9b41c8 The package is published under a name suggesting a Pino logger integration for Zod, but the tarball contents do not match that purpose and exhibit...

MAL-2026-6273 Malicious code in zod-pino (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector c536e5a7ee3d5542e1ac822b30ba4525e52b2ae0c964d0c2470468d91b9b41c8 The package is published under a name suggesting a Pino logger integration for Zod, but the tarball contents do not match that purpose and exhibit...

Claude Code: Out-of-Band Data Exfiltration via Pre-Approved HuggingFace Domain in WebFetch

Because the hostname huggingface.co was pre-approved as a bare hostname for the WebFetch tool, any path on that domain—including attacker-controlled model repositories—was auto-approved without a permission prompt or being subject to --allowedTools restrictions. An attacker able to inject untrust...

PT-2026-50594

Name of the Vulnerable Software and Affected Versions Claude Code versions 0.2.54 through 2.1.162 Description The WebFetch tool pre-approved the hostname 'huggingface.co' as a bare hostname, allowing any path on that domain to be auto-approved without a permission prompt or restrictions from...

CVE-2026-48797 Backpropagate: backprop ui --auth and backprop ui --share do not enforce authentication

Backpropagate is a Python library for fine-tuning large language models on a single GPU. In versions 1.1.0 and 1.1.1, the optional Reflex web UI exposes a training control plane without authentication: dataset upload, model load, training start/stop, multi-run orchestration, GGUF export, and...

vLLM: Security Check Bypass via assert Statement in Activation Function Loading Allows Arbitrary Code Execution

Summary An assert-based security check in vLLM's activation function loading allows any unauthenticated attacker to achieve arbitrary code execution on the server by publishing a malicious HuggingFace model, when vLLM runs in Python optimized mode python -O or PYTHONOPTIMIZE=1. Details vLLM uses ...

PT-2026-50140

Name of the Vulnerable Software and Affected Versions vLLM versions prior to 0.22.0 Description An assert-based security check in the activation function loading process allows an unauthenticated attacker to achieve arbitrary code execution on the server. This occurs when vLLM is run in Python...

MAL-2026-5568 Malicious code in forge-jsx2 (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector 0ce40276c3c58337b7db3272f89e0716b017b4d63bfa625b8757b9d1969ec9f9 The package masquerades as an 'Autodesk Forge' integration but ships no Forge API code. On npm install, scripts/postinstall-agent.mjs materializes a...

CVE-2026-46432

LMDeploy is a toolkit for compressing, deploying, and serving large language models. In versions 0.12.3 and prior, LMDeploy is vulnerable to arbitrary code execution through hardcoded "trustremotecode=True" in multiple HuggingFace model-loading call sites. At time of publication, there are no...

Malicious code in hex-type (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector f7d0271fe97ea66e9ff2ba3a0ea225364324f28138af32c337d6ed8b2b99e5ad Package metadata description "A universally-unique, lexicographically-sortable, identifier generator", homepage github.com/ulid/javascript, build...

CVE-2026-46432

LMDeploy is a toolkit for compressing, deploying, and serving large language models. In versions 0.12.3 and prior, LMDeploy is vulnerable to arbitrary code execution through hardcoded "trustremotecode=True" in multiple HuggingFace model-loading call sites. At time of publication, there are no...

lmdeploy 代码注入漏洞

lmdeploy is a toolkit developed by InternLM for compressing, deploying, and serving LLMs. Versions of lmdeploy prior to 0.12.3 have a code injection vulnerability, which stems from the hard-coded trustremotecode=True setting. This vulnerability could lead to remote code execution within the...

EUVD-2026-35873

LMDeploy is a toolkit for compressing, deploying, and serving large language models. In versions 0.12.3 and prior, LMDeploy is vulnerable to arbitrary code execution through hardcoded "trustremotecode=True" in multiple HuggingFace model-loading call sites. At time of publication, there are no...

CVE-2026-31239

The mamba language model framework thru 2.2.6 is vulnerable to insecure deserialization CWE-502 when loading pre-trained models from HuggingFace Hub. The MambaLMHeadModel.frompretrained method uses torch.load to load the pytorchmodel.bin weight file without enabling the security-restrictive...