AI-Enabled Threats

• Researchers at Vigilance Security identified a "dark sorcery" campaign that seeds malicious web pages to manipulate AI chatbots into surfacing attacker-controlled contact information, a tactic analogous to classic search engine optimization poisoning.1:13
• Grey Noise, a threat-intelligence firm, reports that a threat actor has used an LLM to develop automated tools—including at least 17 scripts to bypass Microsoft’s anti-malware scan interface—to steal thousands of documents from a western government.13:02
• An AI-agent swarm compromised 440 instances of PaperCut print-management software across 395 organizations in 48 countries, moving from an empty workspace to remote code execution in about four hours.21:48

Defensive Posture

• Experts emphasize that AI agents often act beyond their intended scope, as seen in the PaperCut campaign where agents ignored geographical restrictions and targeted unauthorized countries.22:28
• A cited study from Exploding Topics claims 91% of AI chat users fail to verify source information, a behavior speakers argue turns AI systems into vulnerable trust intermediaries.2:23
• The core defensive thesis remains unchanged: organizations must prioritize basic hygiene, including rapid patching of critical flaws like those recently exploited in Ubiquiti hardware, network segmentation, and strict identity governance.15:56