EU Regulatory Changes

This publication is a Systematization of Knowledge (SoK) paper from arXiv that surveys how artificial intelligence is being used to automate binary code reverse engineering. It maps current AI tech...

This publication introduces OTRO, a novel cryptographic protocol for Oblivious Tokenization Path with Square-Root ORAM, designed to enhance privacy and security in data retrieval systems. The frame...

This publication introduces the ARVO framework, a comprehensive atlas cataloguing reproducible vulnerabilities in open-source software components. It systematically documents known security flaws w...

A new preprint from arXiv, titled "Syntactic Systems Cannot See Semantic Invariants," has been published under the AI Safety framework. The paper argues that current large language models and other...

This paper, published on arXiv, presents a novel framework for training autonomous cyber agents using a neurosymbolic approach that learns from observations rather than explicit programming. The re...

This publication introduces a new consensus protocol called Gatling, designed to achieve rapid transaction finality through parallel processing. While not a regulatory change itself, it signals a s...

This paper, published on arXiv, presents a new class of security vulnerability specifically targeting AI agents that use multimodal inputs—such as images, text, and audio. The authors demonstrate t...

A new red-team study published on arXiv evaluates the safety of Anthropic’s Fable 5 and Opus 4.8 models, focusing on their susceptibility to generating harmful or deceptive outputs. The research sy...

A new research paper published on arXiv presents a dataset of multi-source cybersecurity logs labeled with the MITRE ATT&CK framework, along with an evaluation framework for small language models (...