As artificial intelligence agents increasingly transition into physical machinery, ensuring safety before real-world deployment has become a critical engineering challenge.
“Imagine if something like that happened in the physical world,” Howie Sun, CEO and co-founder of Laplace Intelligence, noted. “AI agents are moving from computers into the real world, and we don’t really want physical AI to fail in the real world.”
Ann Arbor-based Laplace focuses on automated red-teaming, which involves stress-testing a system by deliberately attempting to trigger failures. The startup is betting that rigorous AI-driven testing is essential for physical safety.
“Our goal is to help companies discover and fix problems before deployment — and, over time, build a shared foundation of evidence that customers, insurers, and regulators can trust,” Sun said.
Laplace is one of 20 startups competing in Startup Battlefield 200, the startup pitch competition taking place at TechCrunch Disrupt in San Francisco this week, October 13-15. It will present on the Disrupt Stage Tuesday afternoon, October 13.
Image Credits:Laplace Intelligence
Academic Foundations and Research
Sun and co-founder and chief scientist Henry Liu launched Laplace after conducting academic research at the University of Michigan. Sun spent years investigating how to challenge autonomous driving AI in simulation, culminating in a 2023 scientific publication that drew interest from autonomous vehicle developers.
“I realized that I wanted to build this and take it from purely academic research into a product that’s useful in the real world,” Sun said.
The company takes its name from “Laplace’s demon,” an 1814 thought experiment by French scholar Pierre-Simon Laplace regarding predictability based on the exact state of physical systems.
“We are essentially using simulation to predict the real world’s future,” Sun said.
High-Fidelity Simulation Environments
The core of the platform is a high-fidelity simulation reconstructed from real-world geography, paired with an AI agent that continually probes for vulnerabilities in robots, autonomous vehicles, or smart city infrastructure.
The simulation utilizes Nvidia’s Cosmos world simulation platform. By blending map data and references such as Google Street View, Laplace can reconstruct environments like highway systems or pedestrian crossings accurately to scale.
Developers can adjust settings in real time using language prompts, such as adding traffic conditions derived from navigation data providers like TomTom.
The environment features interactive participants. For example, if a test agent vehicle swerves into oncoming traffic, surrounding virtual vehicles react dynamically to avoid a collision.
“It’s truly an interactive simulation environment,” Sun said. “Everyone is reacting to everything … This is more like a world model with interactive participants.”
Automating Edge Case Generation
A key differentiator for Laplace is its use of AI agents to automatically generate complex corner cases—rare, challenging scenarios that could cause a physical AI system to fail—reducing the need for manual human input.
“They can design from the news, from police reports on collisions — all kinds of information to actively challenge the autonomous vehicles and other physical AI agents,” Sun said.
Image Credits:Laplace Intelligence
Market Adoption and Commercialization
While traditional autonomous vehicle developers often rely on recorded footage or manual scenario design, the validation of generalized robotics presents an expanding market as companies deploy humanoid robots into unstructured environments.
“There’s less data on generalized robotics,” Sun said. “No one knows what the edge cases will be in the real world.”
Despite its academic origins, Laplace has secured revenue and achieved profitability. Current customers include autonomous trucking firm Bot Auto and an undisclosed major automotive manufacturer.
The startup is also collaborating with Nvidia and T-Mobile on a smart city traffic pilot for San Jose, California, alongside early-stage projects with developers building humanoid warehouse robots.
Competitive Landscape
Most of Laplace’s contracts remain project-based as the firm transitions toward recurring revenue models. The company faces competition from well-funded sector rivals like Applied Intuition, which has raised substantial capital.
Chief among its rivals is Applied Intuition, a buzzy late-stage startup that’s raised $1.2 billion to date at a $15 billion post-money valuation.
Sun remains confident that the core methodology of utilizing AI to test physical AI represents a paradigm shift toward dynamic interactive testing.
“That will be the biggest moat,” he said.
This article has been updated to clarify Laplace’s relationship with Nvidia.
