CVE-2026-7301 CVE-2026-7301

SGLangs multimodal generation runtime scheduler's ROUTER socket binds to 0.0.0.0 by default and contains a sink that calls pickle.loads on incoming messages, enabling RCE when exposed to the internet...

CVE-2026-7301

SGLangs multimodal generation runtime scheduler's ROUTER socket binds to 0.0.0.0 by default and contains a sink that calls pickle.loads on incoming messages, enabling RCE when exposed to the internet...

CVE-2026-7301 CVE-2026-7301

SGLangs multimodal generation runtime scheduler's ROUTER socket binds to 0.0.0.0 by default and contains a sink that calls pickle.loads on incoming messages, enabling RCE when exposed to the internet...

Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models

Unified autoregressive models UAMs are transformer models that generate text as well as image tokens within a single autoregressive pass. Shared parameters and a multimodal vocabulary simplify the training pipeline and facilitate flexible multimodal generation, yet might introduce new...

SGLang contains two remote code execution and one path traversal vulnerability

Overview Three vulnerabilities have been discovered in the SGLang project, two enabling remote code execution RCE, and one regarding a path traversal vulnerability. In order for an attacker to exploit these vulnerabilities, the multimodal generation mode must be enabled, and an attacker must have...

PT-2026-41670

Name of the Vulnerable Software and Affected Versions SGLang affected versions not specified Description The multimodal generation runtime allows unauthenticated remote code execution when the --enable-custom-logit-processor option is active. This occurs because Python objects loaded through the...

CVE-2026-7304: Deserialization of Untrusted Data

SGLangs multimodal generation runtime is vulnerable to unauthenticated remote code execution when the --enable-custom-logit-processor option is enabled, as Python objects loaded via dill.loads will be deserialized without validation...

PT-2026-41669

Name of the Vulnerable Software and Affected Versions SGLangs multimodal generation runtime affected versions not specified Description An unauthenticated path traversal flaw allows an attacker to write arbitrary files to any location where the server process has write permissions. This is achiev...

Hallucination As Exploit: Evidence-Carrying Multimodal Agents

Multimodal agents use screenshots, documents, and webpages to choose tool calls. When a false visual claim triggers a click, email, extraction, or transfer, hallucination becomes an authorization failure rather than an answer-quality error. We formalize this failure mode as hallucination-to-actio...

CVE-2026-7301: Deserialization of Untrusted Data

SGLangs multimodal generation runtime scheduler's ROUTER socket binds to 0.0.0.0 by default and contains a sink that calls pickle.loads on incoming messages, enabling RCE when exposed to the internet...

CVE-2026-7302: Path Traversal: '.../...//'

SGLangs multimodal generation runtime is vulnerable to an unauthenticated path traversal vulnerability, allowing an attacker to write arbitrary files anywhere the server process has write access, by including ../ sequences in the upload filename when sent to specific endpoints...

sglang 安全漏洞

SGLang is a programming language and runtime system developed by SGL-project, aimed at accelerating large model inference. SGLang has a security vulnerability, which stems from unvalidated path traversal in the multimodal generation runtime. This vulnerability could allow attackers to send files ...

CVE-2026-44222

vLLM is an inference and serving engine for large language models LLMs. From 0.6.1 to before 0.20.0, there is a a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder...

CVE-2026-44222 vLLM: Remote DoS via Special-Token Placeholders

vLLM is an inference and serving engine for large language models LLMs. From 0.6.1 to before 0.20.0, there is a a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder...

CVE-2026-44222

vLLM is an inference and serving engine for large language models LLMs. From 0.6.1 to before 0.20.0, there is a a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder...

CVE-2026-44222

CVE-2026-44222 (vLLM) affects vLLM versions 0.6.1 through 0.19.x where a token-injection vulnerability in multimodal processing allows unauthenticated text prompts containing special tokens to be interpreted as control. When image/video placeholder sequences are provided without corresponding dat...

CVE-2026-44222 vLLM: Remote DoS via Special-Token Placeholders

vLLM is an inference and serving engine for large language models LLMs. From 0.6.1 to before 0.20.0, there is a a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder...

CVE-2026-44222 vLLM: Remote DoS via Special-Token Placeholders

vLLM is an inference and serving engine for large language models LLMs. From 0.6.1 to before 0.20.0, there is a a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder...

The vulnerability of the _download_image_to_temp() function in the system, which is used for launching and managing large language models in LoLLMS (Lord of Large Language Multimodal Systems), allows a hacker to perform an SSRF attack.

The vulnerability of the downloadimagetotemp function in the system for launching and managing large language multimodal systems LoLLMS is related to insufficient testing of requests on the server side. Exploiting this vulnerability could allow a malicious actor to perform an SSRF attack...

CVE-2026-44222: Improper Validation of Array Index

vLLM is an inference and serving engine for large language models LLMs. From 0.6.1 to before 0.20.0, there is a a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder...