The Risk You Didn't Plan For: How Traits Shape Threats - Threats & Vulnerabilities Summit 2025

This 22-minute conference talk from the Threats & Vulnerabilities Summit 2025 explores how agentic systems create new security challenges for organizations. Discover the most significant threats in AI-driven SaaS offerings, including prompt injection vulnerabilities, agent identity abuse, weaknesses in anomaly detection systems, and data poisoning attacks. Learn practical approaches for security teams to identify risks by asking targeted questions, effectively interpreting ambiguous vendor responses, and strengthening internal risk assessment methodologies. Presented by Nate Lee, Founder and CISO of Cloudsec AI, this session provides actionable takeaways to help security professionals better evaluate and mitigate emerging AI security threats. Access additional cloud security resources from the Cloud Security Alliance, including their Top Threats to Cloud Computing 2024 Report, research publications, AI Safety Initiative, online training programs, and opportunities to contribute as a research volunteer.