Five critical AI safety sessions to catch at TechCrunch Disrupt 2026

Anthropic Head of Applied AI Cat de Jong will headline the AI Stage to dissect the divide between stalled pilot projects and successful enterprise deployments. Her session, 'What Anthropic Sees When Enterprises Actually Deploy Claude,' offers a practical look at the barriers preventing widespread adoption. Similarly, Okta’s Ric Smith and NanoCo’s Gavriel Cohen will address the 'Agent Security Problem,' focusing on infrastructure-level vulnerabilities that emerge when AI agents are granted autonomous permissions.

For those targeting enterprise clients, the session 'Securing the AI Enterprise' features AWS VP Rudy Mitra, Luta Security CEO Katie Moussouris, and cybersecurity expert Wendy Nather. They will examine the complex security and governance requirements that define whether a product successfully clears the enterprise procurement process. Meanwhile, on the Real World AI Stage, leaders from Shield AI, General Motors, and Waabi will tackle the high-stakes reality of deploying systems where failure carries physical consequences, such as in defense and aviation.

Finally, NVIDIA’s Les Karpas will lead a discussion on the data limitations hindering robotics development. His session explores how foundation models and simulation environments might finally provide the 'ChatGPT moment' for physical AI. These conversations, part of a broader agenda featuring over 200 sessions, will take place October 13–15 at Moscone West in San Francisco, providing a roadmap for founders aiming to build systems that earn user trust.