AI Sec Watch

Recommender systems (platforms that suggest products or services to users) are vulnerable to data poisoning attacks (malicious manipulation of the data the system learns from to make it behave incorrectly). This paper presents METT, a detection method that identifies these attacks even when they are carefully hidden or small-scale, using techniques like causality inference (analyzing cause-and-effect relationships in user behavior) and a disturbance tolerance mechanism (a way to distinguish real attack patterns from false alarms).

Codex CLI (a coding tool from OpenAI that runs on your computer) versions 0.2.0 to 0.38.0 had a sandbox bug that allowed the AI model to trick the system into writing files and running commands outside the intended workspace folder. The sandbox (a restricted area meant to contain the tool's actions) wasn't properly checking where it should allow file access, which bypassed security boundaries, though network restrictions still worked.

Flowise is a tool with a visual interface for building customized AI workflows. Before August 2025, free-tier users on Flowise Cloud could access sensitive secrets (like API keys for OpenAI, AWS, and Google Cloud) belonging to other users through a Custom JavaScript Function node, exposing data across different user accounts. This cross-tenant data exposure vulnerability has been patched in the August 2025 update.

Flowise version 3.0.5 has a remote code execution (RCE, where an attacker can run commands on a system they don't own) vulnerability in its CustomMCP node. When users input configuration settings, the software unsafely executes the input as JavaScript code using the Function() constructor without checking if it's safe, allowing attackers to access dangerous system functions like running programs or reading files.

Flowise version 3.0.5 contains a Server-Side Request Forgery vulnerability (SSRF, a flaw that lets attackers trick the server into making requests to internal networks on their behalf) in the /api/v1/fetch-links endpoint, allowing attackers to use the Flowise server as a proxy to access and explore internal web services. This vulnerability was patched in version 3.0.6.

A vulnerability (CVE-2025-10772) was found in huggingface LeRobot versions up to 0.3.3 in the ZeroMQ Socket Handler (a tool for sending messages between programs), which allows attackers to bypass authentication (verification of who you are) when accessing the system from within a local network. The vendor was notified but did not respond with a fix.

A vulnerability in Keras (a machine learning library) allows attackers to run arbitrary code on a system by creating a malicious .keras model file that tricks the load_model function into disabling its safety protections, even when safe_mode is enabled. The attacker does this by embedding a command in the model's configuration file that turns off safe mode, then hiding executable code in a Lambda layer (a Keras feature that can contain custom Python code), allowing the malicious code to run when the model is loaded.

A vulnerability exists in Keras' Model.load_model method where specially crafted .h5 or .hdf5 model files (archive formats that store trained AI models) can execute arbitrary code on a system, even when safe_mode is enabled to prevent this. The attack works by embedding malicious pickled code (serialized Python code) in a Lambda layer, a Keras feature that allows custom Python functions, which bypasses the intended security protection.

Lobe Chat, an open-source AI chat framework, has a cross-site scripting vulnerability (XSS, where attackers inject malicious code into web pages) in versions before 1.129.4. When the app renders certain chat messages containing SVG images, it uses a method called dangerouslySetInnerHTML that doesn't filter the content, allowing attackers who can inject code into chat messages (through malicious websites, compromised servers, or tool integrations) to potentially run commands on the user's computer.