Laplace Intelligence uses AI agents to stress-test autonomous vehicles

Laplace Intelligence, a startup originating from the University of Michigan, is developing automated red-teaming technology to ensure the safety of robotics and physical AI systems. By using AI agents to test other AI systems within highly realistic simulated environments, the company aims to identify failures and edge cases before hardware is deployed in the real world.

The company’s simulation technology reconstructs real-world scenarios—ranging from highways in Arizona to pedestrian crossings in Paris—and dynamically adjusts them using real-time traffic and congestion data. Developers can use natural language commands to alter testing conditions, such as increasing traffic flow. The system also simulates the behavior of other road users to ensure logical interactions during stress tests.

Founded by CEO Howie Sun and Chief Scientist Henry Liu, Laplace Intelligence seeks to provide a foundation of evidence that customers, insurers, and regulators can trust. The startup currently serves clients including Isuzu and the Texas-based autonomous trucking company Bot Auto. Laplace is also participating in the TechCrunch Disrupt Startup Battlefield competition.