Shield AI, Waabi, and General Motors on building AI when failure is not an option at TechCrunch Disrupt 2026

Artificial intelligence will be examined through a high-stakes lens at TechCrunch Disrupt 2026, where executives from Shield AI, Waabi and General Motors are scheduled to discuss how developers approach systems that cannot afford serious mistakes. The session, set for the event’s Real World AI Stage, brings together leaders working in distinct corners of automation and robotics. Its focus is not simply on making AI more capable, but on the engineering, judgment and organizational choices involved when reliability has unusually large consequences.

Nathan Michael, Shield AI’s chief technology officer, is due to appear alongside Raquel Urtasun, the founder and chief executive of Waabi, and Mikell Taylor, General Motors’ director of robotics strategy. Their panel is titled “Building AI Systems When Failure Is Not an Option.” By putting those three roles on one stage, the program frames the discussion around a practical question: what changes when an AI system must operate in settings where errors cannot be treated as minor inconveniences?

The subject lands as AI moves beyond software demonstrations and into products, machinery and operational decisions. In lower-risk uses, an incorrect output may be noticed, revised or ignored. In more demanding applications, however, teams have far less room to rely on that kind of recovery. The forthcoming discussion is therefore positioned around the gap between a system that can produce impressive results and one that can be trusted under real-world constraints. That distinction is central to the panel’s premise.

Shield AI, Waabi and General Motors represent different points of view on that problem. One contributor comes from a company led by a technology chief, another from a founder-CEO building an AI business, and the third from a major automaker’s robotics strategy function. Their responsibilities are not interchangeable, which could make the conversation useful for showing how technical development, company leadership and industrial strategy each shape decisions about dependable AI.

There is also an important caveat in the phrase “failure is not an option.” No complex system can be made immune to every possible problem, and the relevant question is often how a team anticipates limits, tests assumptions and responds when conditions depart from expectations. The session’s value will depend on whether it distinguishes ambition from evidence and addresses the trade-offs behind dependable deployment. It may also reveal where participants agree on safeguards and where their organizational priorities lead to different answers.

The panel is part of a broader Disrupt 2026 program built around six industry-focused stages. The event’s advertised lineup includes OpenAI, Anthropic and Replit, among additional participants, placing the real-world AI discussion within a larger gathering of companies and executives involved in the technology sector. That broader format gives attendees a way to compare conversations about AI reliability with sessions centered on other parts of the industry, rather than treating the topic as isolated from the business environment around it.

For the general audience, the larger issue is how promises about artificial intelligence translate into systems people and organizations may eventually depend on. Developers are increasingly judged not only on what their models can do in favorable conditions, but also on the care applied to their boundaries and use cases. The upcoming session cannot resolve that challenge on its own. It can, however, offer a public forum for leaders in AI, robotics and automotive strategy to explain the standards and choices they consider when stakes are high.

Attention will now turn to the details the speakers provide on the Real World AI Stage and to whether their approaches converge or diverge. The event is also promoting a limited ticket offer: attendees can save as much as $200, while a second ticket is discounted by 50%. Those incentives, along with a 25% reduction advertised for tickets, are scheduled to expire at 11:59 p.m. Pacific Time on September 25. The deadline gives prospective attendees a near-term decision ahead of the 2026 gathering.

Source: TechCrunch