Essential AI Safety Discussions for Founders at TechCrunch Disrupt 2026

In the rapidly evolving landscape of artificial intelligence, ensuring the safety and trustworthiness of AI systems has become paramount. TechCrunch Disrupt 2026, a premier event for innovators and industry leaders, is set to host a series of vital discussions dedicated to this critical topic. Across both the AI Stage and Real World AI Stage, five key sessions will delve into the complexities of AI safety, covering everything from securing AI agents in enterprise environments to building autonomous systems where failure is not an option. These sessions offer invaluable insights for founders aiming to develop AI solutions that not only push technological boundaries but also inspire confidence and broad adoption.

As AI transitions from conceptual demonstrations to integral components of businesses and daily life, the focus on its responsible deployment intensifies. Founders are increasingly confronting questions about the reliability of AI agents managing company systems, the safety of autonomous vehicles transporting employees, and the dependable operation of robots in unpredictable physical spaces. These are no longer hypothetical concerns but practical challenges demanding robust solutions. The TechCrunch Disrupt event provides a platform for industry experts to share their experiences and strategies for embedding safety and security into the core of AI products, ensuring they are both innovative and trustworthy.

One of the highlighted sessions, titled 'What Anthropic Sees When Enterprises Actually Deploy Claude,' will feature Cat de Jong, Head of Applied AI at Anthropic. She will discuss the factors that differentiate successful AI deployments from those that falter, drawing from her direct experience with enterprises integrating Claude into their critical workflows. This discussion is particularly relevant for founders seeking to successfully market and deploy their AI solutions within large organizations.

Another crucial session, 'The Agent Security Problem Nobody Is Talking About,' will address the emerging security vulnerabilities introduced by AI agents capable of taking action. Ric Smith, President of Products and Technology at Okta, and Gavriel Cohen, Co-Founder and CEO of NanoCo, will explore infrastructure-level agent security, highlighting the limitations of application-level permissions and essential architectural considerations for deploying agentic AI securely. This session underscores the necessity of prioritizing security from the initial stages of product development.

The session 'Securing the AI Enterprise: Why the Cloud Just Got a Lot More Complicated' will bring together Rudy Mitra, VP of Security Services at AWS, Katie Moussouris, CEO of Luta Security, and cybersecurity veteran Wendy Nather. They will examine the robust infrastructure required for enterprises as AI assumes more autonomous functions. For founders developing AI solutions for corporate clients, understanding these security, governance, and observability requirements is crucial for ensuring successful integration and deployment.

Addressing the profound implications of AI in the physical world, 'Building AI Systems When Failure Is Not an Option' will feature Nathan Michael, CTO of Shield AI; Mikell Taylor, Director of Robotics Strategy at General Motors; and Raquel Urtasun, Founder and CEO of Waabi. This panel will tackle the challenges of determining when autonomous systems are sufficiently safe for deployment, particularly in high-stakes environments like defense, vehicles, and industrial operations. The discussion will cover cultivating a safety-first culture, rigorous testing and validation, and navigating complex regulatory landscapes.

Finally, 'Robots Are Waiting for Their ChatGPT Moment. Here Is What Is Standing in the Way' will delve into the data challenges hindering the scalability of physical AI. Les Karpas, NVIDIA Inception Global Head of Physical AI, will explore how advancements in data pipelines, simulation environments, and foundation models can bridge this gap. This session will provide insights into the steps needed to build more capable and reliable robots and earn the trust required for their widespread adoption in the real world.

These sessions collectively highlight that while building groundbreaking AI models and agents is a significant achievement, the true measure of success lies in the ability to foster trust and facilitate real-world adoption. The insights gained from these discussions at TechCrunch Disrupt 2026 will equip founders with the knowledge and strategies necessary to address the critical safety, security, and reliability challenges, ensuring their innovations have a lasting and positive impact.