AI and Agentic AI promise significant benefits, however, they bring security risks such as breaches of privacy and confidentiality, unauthorized access and action. The unpredictability of Generative AI also poses new risks. We must ensure chat bots, co-pilots and agents stay within their intended scope boundaries and authorizations. Agents that interact with business systems and especially those that can act on users' behalf need to be carefully controlled and these risks multiply when agents start collaborating together.
 
This session gives an overview of the security risks for AI workloads across the stack, from agent and application, down to the models, platforms, and infrastructure. This established cybersecurity practice helps to identify security risks so they can be addressed early. Join this session to learn techniques, tools, and services that SAP uses itself to secure AI workloads and minimize risk without inhibiting pace of innovation.

You will learn:

• Full-stack security risks for AI workloads (SAP or otherwise). 
• Core principles of AI and Agentic AI threat modeling and risk identification. 
• Techniques, tools and services that can help reduce security risks.