The AI-as-weapon story just got a primary source: Amazon Threat Intelligence confirmed a Russian-speaking financially motivated actor used commercial generative AI tools to breach 600+ FortiGate devices across 55 countries between January 11 and February 18. No FortiGate vulnerabilities were exploited — the AI simply made an unsophisticated actor fast and scalable enough to hammer exposed management ports with weak single-factor creds at industrial scale. Amazon CISO CJ Moses called it out directly: this is what happens when AI lowers the barrier to entry for commodity attackers. Separately, Anthropic answered the threat with the same tool — Claude Code Security launched today in limited preview, scanning codebases for vulnerabilities and suggesting patches before adversaries can find them first.
On the exploit front, two CVSS 9.9 critical flaws remain the headline risk. BeyondTrust CVE-2026-1731 is now confirmed by Palo Alto Unit 42 as being weaponized for web shells, C2 installs, lateral movement, and data theft across financial, legal, healthcare, and tech sectors in the US, France, Germany, Australia, and Canada. Roundcube CVE-2025-49113 (also 9.9, deserialization RCE) was added to CISA's KEV catalog today after attackers weaponized it within 48 hours of public disclosure. Both are in active exploitation — patch windows are effectively zero.
Mobile threat landscape deepened with two significant reports. Intellexa's Predator spyware now hooks iOS SpringBoard to suppress the camera and microphone indicator dots while silently streaming to operators — a direct attack on Apple's privacy UI. And PromptSpy, the first confirmed Android malware using Google's Gemini model at runtime to adapt persistence across device configurations, signals that AI-augmented malware is no longer theoretical. Supply chain also took a hit: Cline CLI 2.3.0 was poisoned via a compromised npm token to silently install OpenClaw on developer machines via postinstall script.