ESET’s Tony Anscombe on Shadow AI, ClickFix and Securing AI Agents
,
Michael Novinson (MichaelNovinson) •
August 14, 2026
Tony Anscombe, Chief Security Evangelist, ESET
As organizations embrace agentic artificial intelligence, attackers are increasingly embedding malicious capabilities into downloadable AI skills, creating new avenues for credential theft, code execution and other attacks, said Tony Anscombe, chief security evangelist at Eset. While many AI skills are designed for legitimate purposes, malicious ones can be difficult for users to distinguish before they’re loaded into AI agents.
Digging into recent research from ESET’s
H1 Threat Report,
Anscombe also warned that enterprises risk fueling shadow AI if they respond with blanket bans instead of balanced governance. Organizations should first understand how employees are using AI before implementing policies that encourage responsible adoption while reducing risk.
“You need to audit how AI is being used by your employees. And then you need to put policy in place. The policy needs to be somewhere in the middle. You don’t want to prohibit because if you prohibit, you’ll end up with more shadow AI,” Anscombe said.
In this video interview with ISMG at Black Hat USA 2026, Anscombe also discussed:
Why agentic AI is fundamentally changing how employees interact with enterprise applications;
Why ClickFix attacks continue to succeed by adapting social engineering lures;
How AI behavioral monitoring and managed detection and response can help organizations secure AI agents and conversations.
Worried about AI skills in your organization? ESET’s free
AI Skills Checker
analyzes skill URLs in real time, detecting signs of malicious activity before you install it.
Anscombe has more than 20 years of international leadership experience spanning cybersecurity, privacy, compliance and financial services. A frequent conference speaker and media commentator, he contributes to
WeLiveSecurity
and has also written for other cybersecurity publications.