Using AI agents securely

AI agents take action. They make decisions, call APIs and manage data – often without human supervision. What happens if something goes wrong? This presentation shows how organisations can maintain control without having to forego the benefits of autonomous systems. The focus is on identity-centric security models, in which AI agents are treated as independent, controllable identities with clearly defined rights, traceable actions and controlled access. Participants will gain concrete insights into how typical risks – uncontrolled tool usage, lack of transparency and excessive access rights – can be specifically addressed, and how secure automation simultaneously creates scalable added value for the company.

Agenda

• From LLM to autonomous AI agents: New capabilities, new risks and changing security requirements
• Identity for machines: AI agents as controllable identities with lifecycle, delegation and clear responsibilities
• Security architecture for Agentic AI: Fine-grained authorisation, policy enforcement and context-based access control
• Transparency and control: Monitoring, auditability and targeted intervention options – from kill switch to approval flow
• Scalable integration: Secure integration of AI agents into existing IAM and enterprise architectures for genuine business value

Key question of the webinar: How can autonomous AI agents be managed and secured in such a way that they deliver genuine added value – without compromising control, traceability and governance?

Speakers

Sebastian Rohr has been an expert in PAM architecture and identity management for over 20 years. He develops strategies for the digital process transformation of multinational corporations and companies subject to regulatory requirements (KRITIS, DORA, etc.)

Michael Pattison , Principal Solutions Engineer & CIAM expert with over 20 years’ experience in pre-sales. Specialist in secure, scalable identity solutions using Okta and Auth0.