Jonathan Jaffe, CISO at Lemonade, has built what he predicts will be "the perfect AI system" using agent orchestration to automate vulnerability management at machine speed, eliminating the developer burden of false positive security alerts. His unconventional approach to security combines lessons learned from practicing law against major tech companies with a systematic strategy for partnering with security startups to access cutting-edge technology years before competitors. Jonathan tells David a story that showcases how even well-intentioned people will exploit systems if they believe they won't get caught or cause harm, which has shaped his approach to insider threat detection and the importance of maintaining skeptical oversight of automated security controls. His team leverages AI agents that automatically analyze GitHub Dependabot vulnerabilities, determine actual exploitability by examining entire code repositories, and either dismiss false positives or generate proof-of-concept explanations for developers. Topics discussed: The evolution from traditional security approaches to AI-powered agent orchestration that operates at machine speed to eliminate false positive vulnerability alerts. Strategic partnerships with security startups as design partners, trading feedback and data for free access to cutting-edge technology while helping shape market-ready products. Policy-based security enforcement for cloud-native environments that prevents the need to manage individual pods, containers, or microservices through automated compliance checks. How legal experience prosecuting tech companies provides unique insights into adversarial thinking and the psychology behind insider threats and system exploitation. Implementation of AI vulnerability management systems that automatically ingest CVEs, analyze code repositories for exploitable methods, and generate proof-of-concept explanations for developers. Risk management strategies for adopting startup technology by starting small in non-impactful areas and gradually building trust through demonstrated value and reliability. Transforming security operations from reactive vulnerability patching to proactive automated threat prevention through intelligent agent-based systems. Key Takeaways:  Implement policy-based security enforcement for cloud environments to automate compliance across all deployments rather than managing individual pods or containers manually. Partner with security startups as design partners by trading feedback data for free access to cutting-edge technology while helping them develop market-ready products. Build AI agent orchestration platforms that automatically ingest GitHub Dependabot CVEs, analyze code repositories for exploitable methods, and dismiss false positive vulnerability alerts. Begin startup technology adoption in low-risk or non-impactful areas to build trust and demonstrate value before expanding to critical security functions. Establish relationships with venture capital communities to gain early access to portfolio companies and emerging security technologies before mainstream adoption. Apply healthy skepticism to security controls by recognizing that even well-intentioned employees may exploit systems if they believe they won't cause harm or get caught. Focus AI development efforts on automating time-intensive security tasks that typically require many days of manual developer work into machine-speed operations. Evaluate business risk first before pursuing legal or compliance actions by calculating whether the effort investment justifies potential outcomes and settlements. Listen to more episodes:  Apple  Spotify  YouTube Website