AI Sec Watch

The Moodle OpenAI Chat Block plugin version 3.0.1 has an IDOR vulnerability (insecure direct object reference, where a user can access resources by directly requesting them without proper permission checks). An authenticated student can bypass validation of the blockId parameter in the plugin's API and impersonate another user's block, such as an administrator's block, allowing them to execute queries with that block's settings, expose sensitive information, and potentially misuse API resources.

This paper presents Co-AttenDWG, a new method for detecting offensive content by combining text and images together. The approach uses coattention (a technique where two types of data pay attention to each other simultaneously), dimension-wise gating (a mechanism that selectively emphasizes important features at a detailed level), and expert fusion (combining predictions from multiple specialized models) to better understand how text and visual information relate to each other.

CVE-2025-49655 is a vulnerability in Keras (a machine learning framework) versions 3.11.0 through 3.11.2 where deserialization (converting saved data back into usable form) of untrusted data can allow malicious code to run on a user's computer when they load a specially crafted Keras file, even if safe mode is enabled. This vulnerability affects both locally stored and remotely downloaded files.

CVE-2025-62356 is a path traversal vulnerability (a flaw that lets attackers access files outside intended directories) in all versions of Qodo Gen IDE that allows attackers to read any local files on a user's computer, both inside and outside their projects. The vulnerability can be exploited directly or through indirect prompt injection (tricking the AI by hiding malicious instructions in its input).

CVE-2025-62353 is a path traversal vulnerability (a flaw that lets attackers access files outside intended directories) in all versions of Windsurf IDE that allows attackers to read and write any files on a user's computer. The vulnerability can be exploited directly or through indirect prompt injection (tricking the AI by hiding malicious instructions in its input).

The Senate introduced the AI LEAD Act, which would make AI companies legally liable for harms their systems cause, similar to how traditional product liability (the legal responsibility companies have when their products injure people) works for other products. The act would clarify that AI systems count as products subject to liability and would hold companies accountable if they failed to exercise reasonable care in designing the system, providing warnings, or if they sold a defective system. Additionally, China announced new export controls on rare earth metals (elements essential to semiconductors and AI hardware), which could disrupt global AI supply chains if strictly enforced.

This research presents a federated learning (FL, a technique where multiple parties train an AI model together without sharing raw data) approach for learning the structure of dynamic Bayesian networks (DBN, a statistical model that represents relationships between variables over time) from distributed time series data. The method addresses challenges like data privacy and heterogeneity (when different parties' data follows different patterns), and provides mathematical proof that the approach reliably converges to good results, which hadn't been formally guaranteed before in this setting.

LlamaIndex v0.14.5 is a release that fixes multiple bugs and adds new features across its ecosystem of AI/LLM tools. Changes include fixing duplicate node positions in documents, improving streaming functionality with AI providers like Anthropic and OpenAI, adding support for new AI models, and enhancing vector storage (database systems that store AI embeddings, which are numerical representations of text meaning) capabilities. The release also introduces new integrations, such as Sglang LLM support and SignNow MCP (model context protocol, a standard for connecting AI tools) tools.

ATLAS Data v5.0.0 introduces a new "Technique Maturity" field that categorizes AI attack techniques based on evidence level, ranging from feasible (proven in research) to realized (used in actual attacks). The release adds 11 new techniques covering AI agent attacks like context poisoning (injecting false information into an AI system's memory), credential theft from AI configurations, and prompt injection (tricking an AI by hiding malicious instructions in its input), plus updates to existing techniques and case studies.