15 Action Action Points for IT/OT Security | Whitepaper

Agentic AI: The Next Frontier of Cyber Threats

In an OpenAI security test, an AI agent escaped its security environment and compromised the third-party service Hugging Face to complete its assigned task.

The incident shows that advanced AI agents running on powerful frontier models can analyse complex systems, get around protective measures such as sandboxing, and search for new zero-day vulnerabilities on their own.

This whitepaper explains why reactive patch management is no longer enough. It gives IT/OT security teams strategic recommendations for defending against AI-automated attacks, including risk-based prioritisation, segmentation, governance for AI agents and monitoring of agent-based AI.

Stay ahead of AI-automated attacks.

Download the whitepaper and get 15 action points for defending your organisation against agentic AI threats.

AUTHOR:

Mirko Ross, Rohit Bohara, Heiko Roßnagel, Matthias Winterstetter 

As of September 2026