CVE-2026-108857: Insertion of Sensitive Information into Log File

Hugging Face Text Embeddings Inference through 1.9.4 contains a cleartext logging vulnerability that exposes the configured apikey because the router's Args struct lacks a redact attribute for it. Attackers with access to router logs, container output, or OTLP telemetry can recover the Bearer tok...

EEF-CVE-2026-101028 Ash.count, Ash.exists and Ash.aggregate skip related resources' read policies in filters and sorts

Summary Incorrect Authorization vulnerability in ash-project ash allows an actor to infer data in related records they cannot read via Ash.count/2, Ash.exists/2 and Ash.aggregate/3. Ash.Actions.Aggregate.run/4 lib/ash/actions/aggregate.ex applied only the root resource's read policy before runnin...

CVE-2026-101028 Ash.count, Ash.exists and Ash.aggregate skip related resources' read policies in filters and sorts

Incorrect Authorization vulnerability in ash-project ash allows an actor to infer data in related records they cannot read via Ash.count/2, Ash.exists/2 and Ash.aggregate/3. Ash.Actions.Aggregate.run/4 lib/ash/actions/aggregate.ex applied only the root resource's read policy before running the...

CVE-2026-101028

The ash-project ash framework (versions 2.6.0 to 3.34.6 ) contains an Incorrect Authorization vulnerability in the Ash.Actions.Aggregate.run/4 function. The root cause is that the aggregate path skips applying read policies for related resources, whereas the standard read path correctly applies t...

PT-2026-108749

Incorrect Authorization vulnerability in ash-project ash allows an actor to infer data in related records they cannot read via Ash.count/2, Ash.exists/2 and Ash.aggregate/3. Ash.Actions.Aggregate.run/4 lib/ash/actions/aggregate.ex applied only the root resource's read policy before running the...

CVE-2026-101028: Incorrect Authorization

Incorrect Authorization vulnerability in ash-project ash allows an actor to infer data in related records they cannot read via Ash.count/2, Ash.exists/2 and Ash.aggregate/3. Ash.Actions.Aggregate.run/4 lib/ash/actions/aggregate.ex applied only the root resource's read policy before running the...

CLEANSTART-2026-AH85401 native inference process that Elasticsearch uses to evaluate uploaded machine learning models accepts a model operation that computes a memory address from an offset supplied inside the model, with...

Security vulnerability affects the sonarqube package. The native inference process that Elasticsearch uses to evaluate uploaded machine learning models accepts a model operation that computes a memory address from an offset supplied inside the model, without validating that the offset stays withi...

CVE-2026-105754

A flaw was found in vLLM. The scale-out inference service fails to properly validate caller-supplied multimodal parameters, such as image geometry and cache identifiers, against the active model's configuration. An authenticated remote attacker can exploit this vulnerability by submitting crafted...

BARE-AI: Bit-Flip Attack Resilience in AI Hardware through Built-In Performance Monitors

Deep Neural Networks DNNs are integral to many safety critical systems, yet they remain highly vulnerable to bit-flip attacks BFAs, where a few memory level perturbations can drastically degrade accuracy. Existing defenses incur significant hardware overhead, depend on retraining, or fail against...

EUVD-2026-92844

vLLM: Scale-out disaggregated multimodal transport trusts caller-supplied features...

CVE-2026-105754

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargsdata field, cache identifiers in the features.mmhashes field, ranges in the...

aleph-alpha-inference (1.0.0), evoke-vllm (>=0.1.0, <=0.1.1) +6 more potentially affected by CVE-2026-105758 via vllm (>=0.24.0, <=0.29.0)

vllm PYPI version =0.24.0, =0.1.0, =2.9.0, =2.3.0, =0.1.10, =0.1.0, =0.1.4 Source cves: CVE-2026-105758 Source advisory: MEND:CVE-2026-105758...

CVE-2026-105755 vLLM: Flash late-interaction scoring caches query embeddings under a caller-controlled request id — cross-request integrity break and induced errors on `/score` and `/rerank`

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, flash late-interaction scoring at the /score and /rerank endpoints derives each worker's querykey value from the caller-controlled X-Request-Id header. A concurrent request that reuses a victim's identifier can...

CVE-2026-105754 vLLM: Scale-out disaggregated multimodal transport trusts caller-supplied features

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargsdata field, cache identifiers in the features.mmhashes field, ranges in the...

CVE-2026-105754 vLLM: Scale-out disaggregated multimodal transport trusts caller-supplied features

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargsdata field, cache identifiers in the features.mmhashes field, ranges in the...

NVIDIA Triton Inference Server Auth Bypass (CVE-2026-24207)

Exploits CVE-2026-24207 in NVIDIA Triton Inference Server = 26.02: unauthenticated access to restricted HTTP APIs including /models and model management endpoints. !/usr/bin/env python3 -- coding: utf-8 -- """NVIDIA Triton Inference Server auth bypass CVE-2026-24207.""" import json from kittysplo...

CVE-2026-105754: Improper Input Validation

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargsdata field, cache identifiers in the features.mmhashes field, ranges in the...

Hidden Risks of Jev: An Empirical Study of Security, Privacy, and Dual Use

Jev turns natural-language questions into typed answers and probabilities with low latency and cost, enabling applications to route requests and select tools. While this interface allows Jev to integrate naturally into application workflows as a decision layer, the security and privacy implicatio...

Static Bootstrap Placement for Encrypted Language Model Decoding

Language models increasingly serve prompts that carry private data, and secure inference under homomorphic encryption lets a client outsource the computation without revealing the prompt. Existing secure inference systems run a forward pass without consuming a token under encryption, and generati...

GHSA-42VR-XJ54-VC7V vulnerabilities

Vulnerabilities for packages: airflow-core, azureml-inference-server-http, openstack-glance-2026.1-fips, nemo-rl-cuda-13.0, datahub-ingestion-fips, openstack-octavia-2025.2, openstack-placement-2026.1-fips, superset, tritonserver-backend-vllm-cuda-13.0, semgrep, airflow, awx, request-1276,...