Chinese LLMs Broaden the Gap Between Attackers & Defenders

This Dark Reading roundup highlights a rapid escalation in attacker capability and target diversity driven by AI and exploited vulnerabilities. New Chinese LLMs and AI-agent attacks (agentjacking, phantom squatting, malicious OpenClaw skills, and exposed AI endpoints) are increasing the attack surface and enabling automated, high‑scale abuse. Multiple high‑impact vulnerabilities are being weaponized in the wild — notably CVE-2026-48558 (SimpleHelp) used to deliver the 'Djinn' stealer — and FortiBleed compromises of Fortinet firewalls are being monetized by ransomware gangs. Cisco CUCM and SD‑W