AI & LLM
AI and LLM covers generative models, agents, prompt injection, data security and new artificial intelligence tools. The collection connects capabilities, limitations, operational risks and their impact on technical work.
Google Gemini Hijacked via Messaging Notifications: The 'Dual Illusion' Attack
SafeBreach researchers have demonstrated how the Google Gemini voice assistant on Android can be hijacked through indirect prompt inje…
CISA to Issue Mandatory AI Security Directive for Federal Agencies by Friday
CISA Acting Director Nick Andersen announced that a Binding Operational Directive (BOD) implementing the new AI Executive Order will b…
TA4922 Targets Europe with New Atlas RAT and AI-Assisted Malware Development
Proofpoint tracks the European expansion of TA4922, a Chinese-speaking cybercrime group deploying the new Atlas RAT, RomulusLoader, an…
AI Agents: Only 11% Secure as 'Lethal Trifecta' Exposes 98% of Market
Adversa AI’s AIRQ Q2 2026 benchmark of 100 commercial agents reveals a 'power-protection inversion': as capabilities increase, defense…
Trump Signs AI Executive Order: 30-Day Voluntary Review for Frontier Models
The executive order establishes a voluntary framework for pre-release government access to advanced AI models, tasking the NSA with mo…
AI Zero-Days and OT Vulnerabilities: ESET’s May 2026 Security Briefing
Tony Anscombe’s latest roundup highlights critical failures in Polish water plants, Google’s discovery of the first AI-generated zero-…
Tuskira Unveils Quell: AI Agent Designed to Mitigate Zero-Days Before Patches Exist
Tuskira has launched Quell, an AI agent that maps attack paths and orchestrates compensating controls to neutralize zero-day threats a…
Cybanetix Launches Managed AI Service: AI-Native MDR Powered by Four-Vendor Stack
Cybanetix has unveiled its Managed AI Service, integrating NOMA, SentinelOne, Microsoft, and Exabeam under a unified 24/7 SOC with a s…
BadBone: Dormant AI Backdoor Evades Six Major Security Defenses
BadBone research demonstrates that backdoors in pre-trained AI models remain invisible until customized, maintaining a 0.10% attack su…
Anthropic Grants ENISA Access to Mythos: A Strategic Shift for EU Cybersecurity
Anthropic is granting ENISA access to its Mythos model for vulnerability discovery. As the first EU entity to join Project Glasswing,…
OpenAI Mandates Hardware-Backed Passkeys for Access to Frontier AI Models
Starting June 1, 2026, OpenAI will require Trusted Access for Cyber (TAC) program members to use hardware-backed passkeys, setting a n…
Insight Launches Managed Exposure Defense to Combat AI-Driven Exploit Speed
Insight consolidates CTEM, enterprise patching, supply chain risk, surge engineering, and XDR into a unified managed service designed…