The tradeoffs between flexibility and safety in AI Agents

Not all AI agent architectures offer equal security protection. 

Independent testing by AI security experts at Lakera compared structured process-guided agents with prompt-driven agents across major threat categories, revealing significant differences in vulnerability to information disclosure, infrastructure attacks, content safety violations, and goal hijacking.

Join Rod Rivera, Rasa DevRel, and Matt Fiedler, Lakera AI Product Manager, to examine how different AI agent architectures perform under security testing and why each impacts security outcomes differently. This online event is ideal for AI architects, security professionals, and business leaders who want to understand the security implications of deploying  conversational AI strategies for customer interactions.

What You'll Learn:

• The specific security vulnerabilities that affect prompt-driven AI agents
• How structured dialogue management blocks exploitation attempts
• The business risks of deploying insecure AI: reputation damage, compliance issues, and financial loss
• Architectural principles for building customer-facing agents that stay secure and on-task