EU Regulatory Changes

This publication introduces a taxonomy and framework for model forensics specifically designed for AI-native wireless networks, which are networks where artificial intelligence is deeply integrated...

This publication from May 2026 introduces a novel vulnerability in large language models, termed MetaBackdoor. The research demonstrates that an attacker can embed a hidden backdoor into an LLM by ...

This publication, a pre-print from arXiv dated May 14, 2026, introduces a new taxonomy and benchmark coverage audit for attacks on large language models (LLMs). It systematically categorises the ty...

This publication introduces Veritas, a novel AI-driven framework designed to automatically detect memory corruption vulnerabilities in compiled binary software. Unlike traditional static analysis t...

This publication, titled PickleFuzzer: A Case Study in Fuzzing for Discrepancies Between Python Pickle Implementations, presents a new automated testing tool designed to find security and reliabili...

This publication, a research paper from arXiv, does not represent a regulatory change but rather a significant analytical study that will inform future regulatory frameworks. The paper presents a l...

A new academic paper published on arXiv, titled WARD: Adversarially Robust Defense of Web Agents Against Prompt Injections, introduces a framework designed to protect autonomous web agents from adv...

This paper, published on arXiv, proposes a new framework for securing advanced AI agents by treating them like operating systems. It argues that current AI safety approaches are insufficient for au...

A new preprint from arXiv, titled "Do Coding Agents Understand Least-Privilege Authorization?" examines the security behavior of AI coding agents when implementing authorization controls. The study...