Currently free during beta - premium features coming soon. Subscribe now to lock in early access.
AI_SAFETY

EU Regulatory Changes

371 changes tracked across 24 compliance frameworks including DORA, NIS2, GDPR, EU AI Act, Cyber Resilience Act, and more.

All DORA NIS2 GDPR CSRD MaRisk ISO27001 EU_AI_ACT CRA DSA DMA eIDAS2 SOC2 PCI_DSS HIPAA ISO42001 AMLD6 PSD3 DATA_ACT GPSR CER EUDR CVE BREACH AI_SAFETY
arXiv: GenTI: Benchmarking LLMs for Autonomous IDPS Rule Generation for Unseen Attacks
arXiv: Towards Worst-case Hardness for Low-Noise LPN
arXiv: PriSrv: Privacy-Enhanced and Highly Usable Service Discovery in Wireless Communications
arXiv: GCD: Garbled, Corrected, Demonstrandum -- Fixing and Proving Go's Extended GCD Implementation
arXiv: SentinelRAG: Synthetic Sentinel Knowledge for RAG Database Copyright Protection
arXiv: An Improved CNN-LSTM Based Intrusion Detection System for IoT Networks
arXiv: Membrane: A Self-Evolving Contrastive Safety Memory for LLM Agent Defense
arXiv: An Embarrassingly Simple Detector for Model Extraction Attacks in Large Language Model API Traffic
arXiv: Hybrid CNN-LSTM Framework for Intelligent Cyber Attack Detection and Prevention in U.S. Critical Digital Infra...
arXiv: Cognitive Threat Intelligence and Explainable Federated Security Analytics for distributed Infrastructure Systems
arXiv: Beyond Waveform Robustness: Robust Feature-Vocoder Adversarial Attacks on Automatic Speech Recognition
arXiv: What If Prompt Injection Never Left? Exploring Cross-Session Stored Prompt Injection in Agentic Systems
This publication, a research paper from arXiv, identifies a new vulnerability in AI agentic systems called cross-session stored prompt injection. Unlike traditional prompt injection attacks that oc...
Read analysis →
arXiv: Preserving Data Privacy in Learning Causal Structure with Fully Homomorphic Encryption
A new research paper published on arXiv proposes a method for learning causal structures from data while preserving privacy using Fully Homomorphic Encryption (FHE). This technique allows organizat...
Read analysis →
arXiv: A-Live: Passive Liveness Detection via Neuromuscular Micro-Motion Signatures on Commodity Sensors
This paper, published on arXiv, introduces a novel passive liveness detection method called A-Live, which uses commodity sensors to identify neuromuscular micro-motion signatures. This technology c...
Read analysis →
arXiv: Bernoulli CUSUM and Bayes-Optimal Detection Ceilings for Trust Fraud in Sparse Rating Networks
This paper, published on arXiv, introduces a new statistical method for detecting fraudulent trust ratings in online platforms, specifically designed for sparse data environments where users have f...
Read analysis →
arXiv: Attention-Augmented LSTMs for Automatic Homophonic Ciphertext Decipherment
This publication, dated June 3, 2026, presents a novel machine learning architecture that combines attention mechanisms with Long Short-Term Memory networks to automatically decipher homophonic cip...
Read analysis →
arXiv: SharedRequest: Privacy-Preserving Model-Agnostic Inference for Large Language Models
This publication, SharedRequest: Privacy-Preserving Model-Agnostic Inference for Large Language Models, introduces a novel cryptographic protocol designed to allow multiple parties to query a large...
Read analysis →
arXiv: From Agent Traces to Trust: Evidence Tracing and Execution Provenance in LLM Agents
This paper, published on arXiv, introduces a technical framework called "Evidence Tracing and Execution Provenance" for Large Language Model (LLM) agents. It proposes methods to systematically reco...
Read analysis →
arXiv: NLLog: Lightweight, Explainable SOC Anomaly Detection via Log-to-Language Rewriting
A new research paper, NLLog, has been published on arXiv proposing a method for anomaly detection in Security Operations Centers (SOCs) that converts raw system logs into natural language descripti...
Read analysis →
arXiv: Sequential Data Poisoning in LLM Post-Training
This paper, published on arXiv, presents a new research finding on a vulnerability in large language models (LLMs) during the post-training phase. It demonstrates a method of sequential data poison...
Read analysis →