Disrupt 2026: Five AI Safety Sessions Founders Should Attend
Five Disrupt 2026 sessions tackle Claude deployment, agent security, cloud controls, autonomous safety and robotics' data gap at Moscone West.
Summary
At TechCrunch Disrupt 2026, five sessions across the AI Stage and Real World AI Stage address safety barriers to deployment. “What Anthropic Sees When Enterprises Actually Deploy Claude” brings Anthropic Head of Applied AI Cat de Jong to explain why some Claude projects produce measurable value while others remain pilots after 18 months. “The Agent Security Problem Nobody Is Talking About” pairs Okta President of Products and Technology Ric Smith with NanoCo co-founder and CEO Gavriel Cohen on infrastructure-level agent controls, application-level permission weaknesses and architecture. “Securing the AI Enterprise: Why the Cloud Just Got a Lot More Complicated” has AWS VP of Security Services Rudy Mitra, Luta Security CEO Katie Moussouris and cybersecurity veteran Wendy Nather covering security, governance and observability as enterprise AI becomes more autonomous.
“Building AI Systems When Failure Is Not an Option” features Shield AI CTO Nathan Michael, General Motors Director of Robotics Strategy Mikell Taylor and Waabi founder and CEO Raquel Urtasun on safety culture, testing, validation, regulation and trust for vehicles, aircraft, industrial systems and critical missions. “Robots Are Waiting for Their ChatGPT Moment. Here Is What Is Standing in the Way” has Nvidia Inception Global Head of Physical AI Les Karpas examine scarce robotics training data and how pipelines, simulation and foundation models could improve capability and reliability. The event runs October 13 to 15 at Moscone West in San Francisco, with 200-plus sessions across six industry stages, roundtables and breakouts; more than 10,000 founders, investors, operators and tech leaders; 250-plus speakers; and 300-plus exhibiting startups. Passes are discounted by up to $200 until September 25 at 11:59 p.m. PT, with a second pass 50% off.
Positives
- Anthropic’s Cat de Jong will identify why some Claude deployments generate measurable enterprise value while others stall.
- Five sessions connect enterprise security, agent permissions, physical autonomy and robotics reliability to real-world AI adoption.
- Nvidia Inception’s Les Karpas will assess data pipelines, simulation and foundation models as potential answers to robotics’ training-data shortage.
- More than 10,000 attendees, 250-plus speakers and 300-plus exhibiting startups are expected across the October 13 to 15 event.
- Passes remain up to $200 cheaper through September 25 at 11:59 p.m. PT, while second passes are 50% off.
Risks & concerns
- Some enterprise AI projects remain pilots after 18 months instead of producing measurable value.
- Application-level permission models may not adequately control AI agents capable of taking actions inside company systems.
- Security, governance and observability become more complicated as enterprise AI assumes increasingly autonomous roles.
- Failures in autonomous vehicles, aircraft, industrial systems and critical missions can cause physical consequences.
- Robotics lacks the massive training-data pools that accelerated language models and self-driving technology.