AI & LLM Vulnerabilities

Langflow versions up to 1.6.9 have a chained vulnerability that allows attackers to take over user accounts and run arbitrary code on the system. The flaw combines two misconfigurations: overly permissive CORS settings (CORS, or cross-origin resource sharing, allows webpages from different domains to access each other) that accept requests from any origin with credentials, and refresh token cookies (a token used to get new access credentials) set to SameSite=None, which allows a malicious webpage to steal valid tokens and impersonate a victim.

NVIDIA Triton Server for Linux has a vulnerability where attackers can bypass input validation (improper validation of specified quantity in input) by sending malformed data. This flaw could allow an attacker to cause a denial of service attack (making a system unavailable to legitimate users).

NVIDIA Triton Inference Server has a vulnerability (CVE-2025-33201) where an attacker can send extremely large data payloads to bypass safety checks, potentially crashing the service and making it unavailable to legitimate users (a denial of service attack). The vulnerability stems from improper validation of unusual or exceptional input conditions.

vLLM (a tool for running large language models) versions before 0.11.1 have a critical security flaw where loading a model configuration can execute malicious code from the internet without the user's permission. An attacker can create a fake model that appears safe but secretly downloads and runs harmful code from another location, even when users try to block remote code by setting trust_remote_code=False (a security setting meant to prevent exactly this).

LibreChat, a ChatGPT alternative with extra features, had a vulnerability in versions before 0.8.1-rc2 where an authenticated user could exploit the "Actions" feature by uploading malicious OpenAPI specs (interface documents that describe how to connect to external services) to perform SSRF (server-side request forgery, where the server itself is tricked into accessing restricted URLs on the attacker's behalf). This could allow attackers to reach sensitive services like cloud metadata endpoints that are normally hidden from regular users.

Keras version 3.11.3 has a path traversal vulnerability (a security flaw where attackers can write files outside the intended directory) in the keras.utils.get_file() function when extracting tar archives (compressed file formats). The function fails to properly validate file paths during extraction, allowing an attacker to write files anywhere on the system, potentially compromising it or executing malicious code.

Fugue is a tool that lets developers run Python, Pandas, and SQL code across distributed computing systems like Spark, Dask, and Ray. Versions 0.9.2 and earlier have a remote code execution vulnerability (RCE, where attackers can run arbitrary code on a victim's machine) in the RPC server because it deserializes untrusted data using cloudpickle.loads() without checking if the data is safe first. An attacker can send malicious serialized Python objects to the server, which will execute on the victim's machine.

LangChain, a framework for building AI agents and applications powered by large language models, has a template injection vulnerability (a security flaw where attackers can hide malicious code in text templates) in versions 0.3.79 and earlier and 1.0.0 through 1.0.6. Attackers can exploit this by crafting malicious template strings that access internal Python object data in ChatPromptTemplate and similar classes, particularly when an application accepts untrusted template input.

Roo Code is an AI-powered coding agent that runs inside code editors. Before version 3.26.7, a validation error allowed Roo to automatically execute commands that weren't on an allow list (a list of approved commands), which is a type of command injection vulnerability (where attackers trick a system into running unintended commands).