When AI controls aircraft, vehicles, or robots, a mistake can cause real-world harm. At TechCrunch Disrupt 2026, executives from Shield AI, Waabi, and General Motors will confront the question of how to build and deploy AI systems when failure is not an option.
TechCrunch Disrupt 2026 will host a panel titled “Building AI Systems When Failure Is Not an Option,” featuring Shield AI chief technology officer Nathan Michael, Waabi founder and CEO Raquel Urtasun, and General Motors’ Director of Robotics Strategy Mikell Taylor. The conversation, which will take place on the Real World AI Stage, is expected to cover the practical work behind deploying artificial intelligence in safety-critical physical environments.
The panelists bring distinct perspectives from defense drones, autonomous trucking, and automotive robotics — sectors where a malfunction can ground an aircraft, cause a vehicle crash, or compromise a mission. The event description notes that an AI chatbot giving a bad answer is one thing, but the stakes rise sharply when AI moves into the physical world.
The session will delve into creating a safety culture, testing and validating autonomous systems, navigating regulatory hurdles, and earning public trust. The discussion aims to provide insights into how organizations determine when an autonomous system is truly ready to operate outside the lab.
Analysis
Why This Matters
- Physical-world AI systems are entering public roads, airspace, and workplaces at an accelerating pace, making safety assurance a pressing regulatory and societal issue.
- The panel signals that major players in defense, autonomous driving, and industrial robotics are coalescing around shared challenges — testing, validation, and trust — rather than competing on secrecy.
- How these companies address failure modes could shape industry standards and influence public acceptance of autonomous technologies.
Background
Autonomous systems have long been tested in controlled environments, but the transition to real-world deployment has been uneven. High-profile incidents involving self-driving vehicles and drone operations have heightened scrutiny from regulators and the public. Companies in this space must demonstrate that their AI can handle edge cases without causing harm, often operating under different regulatory frameworks across jurisdictions.
Key Perspectives
Defense & Aerospace (Shield AI): Nathan Michael brings experience with autonomous aerial systems used in military contexts, where mission reliability and fail-safe design are paramount. The company’s focus is on AI that can operate without GPS or communications in contested environments.
Autonomous Trucking (Waabi): Raquel Urtasun has advocated for a simulation-first approach to validation, arguing that rigorous testing in virtual environments can catch failures before they reach public roads. Waabi’s technology is aimed at long-haul freight.
Automotive Robotics (General Motors): Mikell Taylor represents a legacy automaker investing in both autonomous vehicle technology and broader robotics for manufacturing. GM’s perspective includes scaling production of safe systems and navigating consumer safety expectations.
What to Watch
- Whether the panel introduces concrete metrics or benchmarks for determining AI system readiness.
- Any mention of regulatory proposals or collaborations that could emerge from the discussion.
- How each company’s approach to simulation vs. real-world testing differs and whether a consensus framework is proposed.