Laplace Intelligence red-teams robots before real-world deployment

Laplace Intelligence is building automated red-teaming — deliberately trying to make a system fail — for robotics and physical AI, betting that the only reliable way to vet physical AI before deployment is to have AI test it. “AI agents are moving from computers into the real world, and we don’t really want physical AI to fail in the real world,” CEO and co-founder Howie Sun told TechCrunch.

Sun and co-founder and chief scientist Henry Liu founded the Ann Arbor, Michigan-based startup a year ago out of the University of Michigan, building on research Sun spent eight years pursuing in academia on challenging autonomous-driving AI in simulation. A 2023 paper he published with Liu — then his PhD adviser — made the cover of Nature and drew interest from autonomous-vehicle companies, prompting Sun to turn the research into a product.

At the core of Laplace’s system is a high-fidelity simulation environment, a post-trained version of Nvidia’s Cosmos world-simulation platform, reconstructed from real-world data such as Google Street View imagery and map data, and paired with an AI agent that constantly probes for weaknesses. Developers can change the setting in real time using language prompts, and Laplace pulls live traffic and congestion data from navigation providers like TomTom. Both the environment and the behavior of other actors in it are simulated and interactive, letting the AI agent design rare, tricky “corner cases” — drawn from sources like news reports and police collision reports — that would otherwise be designed by humans.

Laplace is generating revenue and a modest profit, mostly from autonomous-driving customers including Japanese trucking manufacturer Isuzu and Texas-based autonomous trucking firm Bot Auto. Nvidia is also a customer, and the two companies are working with T-Mobile on a free smart-city pilot for San Jose, California, covering high-level traffic operations; Laplace is also in early talks with partners building humanoid robots for warehouses. Most of its contracts are project-based for now, and the startup — one of 20 companies in this week’s Startup Battlefield 200 competition at TechCrunch Disrupt in San Francisco — is looking to raise funds and scale to compete with Applied Intuition, a rival that has raised $1.2 billion to date at a $15 billion post-money valuation and counts Isuzu among its own customers too.

Why it matters

Laplace’s bet is that simulated red-teaming becomes standard practice as robots and autonomous systems move from controlled factory floors into unpredictable public spaces, and that the evidence it generates comes to matter not just to engineering teams but to insurers and regulators too.