Shield AI, Waabi, and General Motors on building AI when failure is not an option at TechCrunch Disrupt 2026
TechCrunch TechCrunch Events ● Covered by 6 sources
Shield AI, Waabi, and GM are talking about AI that can’t fail in the real world. The hard part isn’t demos — it’s proving autonomy won’t wreck an aircraft, truck, or robot.
Based on reporting by TechCrunch, TechCrunch Events — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
TechCrunch Disrupt 2026 is setting up a very specific kind of AI debate: what happens when the machine isn’t just answering prompts, but steering something that can crash, break, or miss its mission. On the Real World AI Stage, Shield AI’s Nathan Michael, Waabi founder and CEO Raquel Urtasun, and General Motors robotics leader Mikell Taylor will talk about building autonomous systems when “good enough” is not remotely good enough.
The session, titled “Building AI Systems When Failure Is Not an Option,” is built around a simple problem with messy consequences. If an AI model is running an aircraft, a vehicle, or a robot, you need more than a clever demo. You need safety culture, testing, validation, regulatory work, and trust. And you need to know when a system has actually earned the right to leave the lab.
Michael brings the defense side of that problem. He leads development and deployment of Hivemind, Shield AI’s platform-agnostic mission autonomy software, after years at Carnegie Mellon University’s Robotics Institute, where he directed the Resilient Intelligent Systems Lab. Hivemind was selected in February as an autonomy provider for the U.S. Air Force’s Collaborative Combat Aircraft drone prototype program, and the company announced $1.5 billion in Series G funding a month later at a $12.7 billion post-money valuation.
Urtasun arrives with a long autonomous-driving résumé and a very unsentimental view of validation. She has spent 25 years in AI and autonomous vehicles, previously led Uber ATG’s research and development, and co-founded the Vector Institute for AI. Waabi raised $1 billion in January and announced a partnership with Uber to support deployment of 25,000 or more Waabi Driver-powered robotaxis, but Urtasun has said its autonomous trucks still need to be fully validated before driverless deployment. Taylor, meanwhile, has spent more than two decades on robots meant to do actual work, from Amazon’s Proteus to industrial systems, with user experience and adoption treated as part of the product, not an afterthought.
The bigger point is plain enough: in defense, trucking, and industrial robotics, the question isn’t whether AI looks impressive. It’s whether people can trust it when the stakes are ugly and the room for error is basically zero. Tech demos are cheap. Assurance is the bill.
My take — AI-written commentary, not fact-checked reporting
This is the useful kind of AI event: less hype theater, more “show the receipts.” The industry keeps pretending autonomy is a product launch when it’s really a long, fussy trust exercise with regulators, users, and physics all in the room. That part is less glamorous than a keynote, which is probably why it matters more.
Read more about this at: TechCrunch