Apple changes full-disk access permissions to curb abuse from AI agents

Ars Technica security feed covering active exploitation of a critical Zimbra command-injection flaw, large-scale government and enterprise data breaches, AI-assisted account compromises, vulnerabilities in privileged AI agents, ClickFix and scareware campaigns, browser and Windows exploit kits, surveillance-camera exposure, and emerging risks involving AI security and cryptography. The most immediately actionable item is exploitation of the Zimbra vulnerability to steal emails; multiple other reports describe active or high-impact attacks and compromises.