When AI Agents Go Rogue & The Threat of Poisoned Models

By 2026, cybercrime will no longer just be a service, it will be a fully automated industry . Trend Micro predicts a seismic shift where autonomous AI agents can scout targets, infiltrate systems, steal data, and even negotiate ransoms without a single human finger on the keyboard .

Sage Khor, Technical Lead at Trend Micro, joins us to break down this AI-fication of cybercrime . He explains the dangers of "machine-speed" attacks hitting organizations that still respond at "human speed" , the rise of "poisoned" AI models that trick systems into ignoring threats , and why Malaysia saw a 29% spike in data breaches in early 2025.

We discuss:

• The 2026 Prediction: How cybercrime moves from "as-a-service" to fully automated factories .

• Machine Speed vs. Human Speed: Why traditional defenses can't keep up with AI agents .

• Poisoned AI Models: How attackers "teach" AI to ignore dangerous threats

• The Cloud Risk: Why simple misconfigurations are often more dangerous than zero-day exploits .

• Resilience Strategy: Why the focus must shift from blocking every attack to rapid containment and "cyber risk exposure management"

Related & Mentioned Podcasts

• Rise of the Agentic Enterprise: AI That Plans & Act

• The 7 Sins of Enterprise AI Investment

• Handling the Hack: The 4 Rules of Crisis Comms 

See omnystudio.com/listener for privacy information.