Every entry below was pulled directly from a regulator’s feed in the last few days — CELLAR, EBA, ESMA, ENISA, BSI, BaFin, ANSSI and 14 other sources — classified against 21 EU compliance frameworks and summarised. Updates daily.
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Anchors that Don't Lift: Understanding Supply Chain Driven Kernel Lock-In and Governance-Mediated Mitigation Strategies in SOHO Devices
This paper, published on arXiv, is not a regulatory change but a research study that identifies a critical supply chain security vulnerability in small office/home office (SOHO) networking devices. The research…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: OpenPCC: Open and Confidential LLM Serving on Commodity TEEs
This paper, published on arXiv, introduces OpenPCC, a technical framework for running large language models (LLMs) on commodity Trusted Execution Environments (TEEs) while maintaining both performance and data…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: A Longitudinal Study of Recently Observed Malicious Domains: Characteristics, Infrastructure, and Abuse Patterns
This publication is a research paper from arXiv, not a regulatory change, but it provides critical empirical evidence that should inform AI safety compliance frameworks. The study analyzes a longitudinal dataset of…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Do Transformers Actually Help Intrusion Detection? A Temporal Sequence Evaluation on CIC-IDS2017
This publication is a research paper, not a regulatory change, but it has significant implications for compliance professionals overseeing AI-driven cybersecurity systems under frameworks like the EU AI Act. The study…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: When Discovery Outpaces Remediation: Modeling AI-Accelerated Vulnerability Discovery in Interconnected Systems
This paper, published on arXiv, models a new systemic risk: AI systems can discover software vulnerabilities far faster than humans or traditional tools can patch them. It demonstrates that in interconnected digital…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Understanding and mitigating the risks of OpenClaw for non-technical users: A practical guide with Skill
This document, published on arXiv, is a practical guide titled "Understanding and mitigating the risks of OpenClaw for non-technical users." It introduces a new risk framework, AI_SAFETY, specifically designed to…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Context-Based Adversarial Attacks on AI Code Generators: Vulnerability Analysis and Implications
This publication, a research paper from arXiv, presents a new vulnerability analysis of AI code generators. It demonstrates that these systems can be manipulated through context-based adversarial attacks, where…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: What Do Deepfake Speech Detectors Actually Hear?
This paper, published on arXiv, presents a technical analysis of deepfake speech detectors, revealing that these systems often rely on superficial acoustic artifacts—such as background noise or recording device…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Ethical and Technical Limits of Deepfake Speech Datasets
This publication from arXiv, dated June 2026, presents a critical analysis of the ethical and technical limitations inherent in current deepfake speech datasets used to train AI systems. While not a regulatory mandate…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: RAT: Reference-Augmented Training for ASV Anti-Spoofing
This publication from arXiv presents a new AI training method called Reference-Augmented Training, or RAT, designed to improve the security of automatic speaker verification systems against spoofing attacks. The…
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Comparative Analysis of Inference-Time Defense Methods for Multimodal Large Language Models
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Training LLMs to Enforce Multi-Level Instruction Hierarchies via Gravity-Weighted Direct Preference Optimization
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Securing Code Understanding: Detecting Natural Backdoor Vulnerability in Code Language Models
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: RedAct: Redacting Agent Capability Traces for Procedural Skill Protection
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: A Bayesian Network Approach for Enhancing Security-Focused Decision Support Systems
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Secure Aggregation with Top-K Sparsification in Decentralized Federated Learning
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Toward Secure LLM Agents: Threat Surfaces, Attacks, Defenses, and Evaluation
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: MemVenom: Triggered Poisoning of Multimodal Memories in Web Agents
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Fingerprinting All AI Cluster I/O Without Mutually Trusted Processors
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Do LLMsMakeNeural Distinguishers Wise?
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: In Defense of Information Leakage in Concept-based Models
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Post-Quantum Secure Federated DeFi for Inclusive Banking
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: Layer Order Semantics for Automata-Based Cybersecurity
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: From Transactions to Records: Reconceptualizing Blockchain Systems through a Lifecycle Lens
AI_SAFETYarxiv_cscr9 Jun 2026
arXiv: snaproot: Decentralized File Integrity Verification Using Blockchain-Anchored Cryptographic Hashing
BREACHransomwarelive9 Jun 2026
Ransomware: worldleaks claims M1xchange (IN) — Technology
BREACHransomwarelive9 Jun 2026
Ransomware: worldleaks claims Apollo Pipes (IN) — Manufacturing
BREACHransomwarelive9 Jun 2026
Ransomware: worldleaks claims GDL Transport (SE) — Transportation/Logistics
BREACHransomwarelive9 Jun 2026
Ransomware: chaos claims airespring.com (US) — Technology
BREACHransomwarelive9 Jun 2026
Ransomware: nova claims Trevi (IT) — Construction