SectionAI Security
Attacks on and with AI: models, agents, data.
Proofpoint to Acquire Varonis in $6.3 Billion All-Cash Deal
The Thoma Bravo-owned email security giant is buying data security specialist Varonis in the largest cybersecurity acquisition announced so far in 2026,…
OpenAI Engineer Arrested Over Hidden Recorder in Executive Conference Room
Saudi national Abdulrahman Alfozan was arrested and fired after a covert listening device was discovered in a San Francisco conference room used…
Google Launches Gemini 4 Argon, Touts Vulnerability Discovery and Patching at Scale
Google DeepMind's new frontier model is now globally available, with the company highlighting its use in finding and fixing critical security flaws…
Report: Most Third-Party AI Agents Are Invisible to Identity Systems
A new State of Agent Security Report finds that roughly 1,000 of 1,280 AI-embedded third-party products studied bypass single sign-on entirely, leaving…
Chinese Hacker Used ARTEX AI and Claude Agents to Breach South Korean Banks
CrowdStrike links a Chinese-speaking threat actor to attacks on Shinhan, KB Kookmin, and Hana Bank, using an AI pentest suite and Claude…
Criminal IP Unveils AITEM, an AI-Driven Push Beyond Traditional Attack Surface Management
AI SPERA is positioning its new AITEM framework as a shift from simply discovering exposed assets to prioritizing and acting on them,…
Anthropic Pulls Live Internet Access From Internal Claude Tests
After internal evaluations showed Claude exploiting injection flaws and acting on real websites, Anthropic has cut off live internet access for all…
OpenAI Fires Three Safety Researchers Amid AI Risk Dispute
OpenAI has dismissed three researchers from its safety team, saying they violated internal policies on handling sensitive information, amid an ongoing disagreement…
OpenAI Agents ‘Escape’ Sandbox, Trigger Wikimedia Outage
Autonomous AI agents reportedly broke out of their intended scope and hammered Wikimedia infrastructure, causing a service outage while also attempting to…
Wikimedia Finds Rogue OpenAI Agents Tried to Weaponize Its Tools as Proxies
Wikimedia says autonomous agents it attributes to OpenAI probed a citation tool and its Etherpad service in apparent attempts to turn them…
Apple to Tighten macOS Full Disk Access Controls as AI Agents Raise Privacy Risks
Apple says it will add new safeguards around Full Disk Access permissions on macOS, warning that increasingly autonomous AI agents could exploit…
Google Pauses Open Source Bug Bounty Over Flood of Invalid AI Reports
Google has halted product vulnerability submissions to its OSS Vulnerability Reward Program after automated, mostly invalid reports overwhelmed the process, joining a…