At TechCrunch Disrupt 2026, a major tech conference in San Francisco, several sessions will focus on the challenges of creating AI systems that are safe, secure, and trustworthy. As AI moves from research labs and demonstrations into real-world use in areas like business operations, self-driving cars, robots, and autonomous systems, ensuring their reliability and security has become a top priority. Founders and developers are now being encouraged to address these concerns early on, so that businesses and users can trust and adopt AI technologies confidently. During the event, Anthropic’s Head of Applied AI, Cat de Jong, will share insights from enterprises using the company’s AI assistant, Claude. She will highlight what makes some AI implementations successful in real-world settings, compared to those that stay in limited testing phases. Other sessions will be led by Ric Smith, President of Products and Technology at Okta, and Gavriel Cohen, Co-Founder and CEO of NanoCo. They will tackle the "agent security problem," which refers to the vulnerabilities in systems that involve autonomous agents—like AI-driven software or robots—by examining weaknesses in how these systems are secured at both the infrastructure and application levels. Another panel will bring together experts in enterprise security, including Rudy Mitra, VP of Security Services at AWS, and Katie Moussouris, CEO of Luta Security, along with cybersecurity veteran Wendy Nather. They will discuss the complex security challenges that come with integrating AI into business environments, especially as AI systems take on more independent roles. Meanwhile, experts like Nathan Michael from Shield AI and Raquel Urtasun from Waabi will focus on the high-stakes nature of AI systems used in critical applications, such as robotics in manufacturing or autonomous vehicles. They will emphasize the importance of developing a strong safety culture, thorough testing, and navigating regulatory requirements to prevent failures that could have serious consequences. In another session, Les Karpas, Inception Global Head of Physical AI at NVIDIA, will address the data challenges involved in training AI systems for physical tasks, such as those used in robotics. He will explore how data pipelines, simulation environments, and foundation models—pre-trained AI systems that can be fine-tuned for specific tasks—can help overcome the lack of training data in these areas. The sessions at TechCrunch Disrupt 2026 aim to bridge the gap between cutting-edge AI research and its practical, real-world application by addressing the key challenges that must be solved before AI can be widely adopted. The event will take place from October 13 to 15 at Moscone West in San Francisco, drawing more than 10,000 attendees.