Improving Methodologies for Agentic Evaluations across Domains: Leakage of Sensitive Information, Fraud and Cybersecurity Threats

The rapid rise of autonomous AI systems and advancements in agent capabilities are introducing new risks due to reduced oversight of real-world interactions. Yet agent testing remains nascent and is still a developing science. As AI agents begin to be deployed globally, it is important that they...