Laplace Intelligence uses AI to red-team robots and autonomous vehicles before real-world deployment

What Laplace Intelligence does

Laplace Intelligence, based in Ann Arbor, is building automated red-teaming for robotics and physical AI systems. The startup uses AI-driven simulations to deliberately stress-test these systems by identifying and exploiting edge cases before they are deployed in the real world. According to TechCrunch, this approach aims to prevent failures in autonomous vehicles, robotics, and other AI-driven technologies by addressing vulnerabilities in a controlled environment.

The company’s simulations are grounded in real-world data, combining sources like Google Street View imagery, map data, and real-time traffic information from providers such as TomTom. This high-fidelity simulation environment is built on Nvidia’s Cosmos platform, enabling accurate and scalable testing of AI agents in scenarios that mimic real-world conditions.

Founders and origins

Laplace Intelligence was founded a year ago by Howie Sun, who serves as CEO, and Henry Liu, the company’s chief scientist. Both founders emerged from the University of Michigan, where they conducted research on autonomous driving AI. Their work gained prominence in 2023 when a paper they co-authored was featured on the cover of Nature, highlighting their contributions to the field of AI safety and simulation.

Sun has spent nearly a decade in academia researching how to challenge autonomous driving AI in simulation. This expertise underpins Laplace’s approach to automated red-teaming, which he describes as a "new paradigm that changes the testing of physical AI entirely."

Customers and revenue

Laplace Intelligence is already generating revenue and has achieved a modest profit. Its customer base includes prominent names in the autonomous vehicle and robotics industries, such as Japanese trucking manufacturer Isuzu, Texas-based autonomous trucking firm Bot Auto, and Nvidia. The startup is also exploring partnerships beyond autonomous driving, including early-stage collaborations with companies developing humanoid robotics for warehouse applications.

In addition to its commercial customers, Laplace is working on a smart city pilot for San Jose, California, in collaboration with Nvidia and T-Mobile. The project, which is offered as a free pilot, aims to demonstrate the potential of AI-driven simulations in urban planning and infrastructure management.

Competition and fundraising goals

Laplace is currently competing in TechCrunch Disrupt’s Startup Battlefield 200, a showcase for early-stage startups. The company’s near-term goal is to raise funds to scale its operations and compete with well-funded rivals in the simulation and testing space. Its most notable competitor, Applied Intuition, has raised $1.2 billion to date and is valued at $15 billion. Isuzu, one of Laplace’s customers, also works with Applied Intuition, highlighting the overlap in their target markets.

The startup’s focus on AI-driven red-teaming sets it apart from traditional simulation providers. By using AI agents to automatically design and test edge cases, Laplace aims to create a more dynamic and adaptive testing framework for physical AI systems.

Why AI safety in robotics matters

The rise of physical AI agents—such as autonomous vehicles, humanoid robots, and smart city infrastructure—has amplified the need for robust testing and safety mechanisms. Unlike digital AI, which can fail without physical consequences, failures in robotics or autonomous systems can pose significant risks to public safety. This has led to increased research and investment in simulation-based testing, where edge cases and potential failures can be identified and addressed before real-world deployment.

Laplace’s approach aligns with broader industry trends, where companies like OpenAI and Hugging Face have emphasized the importance of red-teaming and adversarial testing for AI systems. The challenge lies in creating simulations that accurately reflect the complexity of the real world, a hurdle Laplace addresses by integrating diverse data sources into its platform.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

Laplace Intelligence represents a growing recognition that AI safety isn’t just a software problem—it’s a physical one. For founders and operators in robotics, autonomous vehicles, or smart infrastructure, the startup’s approach offers a blueprint for mitigating risk without relying solely on costly and time-consuming real-world testing.

The fact that Laplace is already generating revenue and working with major customers like Isuzu and Nvidia suggests there’s real demand for scalable, AI-driven testing solutions. However, the competition is fierce, with well-capitalized players like Applied Intuition dominating the space. Laplace’s participation in Startup Battlefield could be a critical opportunity to attract investors and partners who see the value in its automated red-teaming methodology.

For operators, Laplace’s work underscores the importance of simulation as a core part of the development cycle for physical AI. The ability to identify and fix edge cases in a virtual environment could accelerate deployment timelines and reduce the likelihood of costly or dangerous failures. Startups in this space should consider how simulation-driven testing can be integrated into their own product development pipelines, especially as AI agents move from digital environments into the real world.

Key takeaways

• Automated red-teaming and simulation-driven testing are becoming critical for AI safety in robotics and autonomous systems.
• Real-world data integration—such as maps, traffic info, and imagery—enhances the accuracy and utility of AI testing environments.
• Startups like Laplace are proving that revenue and profitability are possible in the AI safety space, even with competition from well-funded incumbents.
• Founders in robotics, autonomous vehicles, or smart infrastructure should prioritize simulation-based testing to identify edge cases before real-world deployment.
• Collaborations with cities, tech giants, and telecom providers (e.g., San Jose, Nvidia, T-Mobile) can open doors for pilots and scaling opportunities.

FAQ

What is red-teaming in the context of AI and robotics?

Red-teaming refers to the practice of deliberately testing a system by trying to make it fail. In AI and robotics, this involves identifying edge cases or vulnerabilities in autonomous systems, such as self-driving cars or robots, before they are deployed in the real world. Laplace Intelligence automates this process using AI-driven simulations.

Why is simulation important for autonomous vehicles and robotics?

Simulation allows companies to test AI systems in a controlled, virtual environment that mimics real-world conditions. This is critical for identifying potential failures or edge cases without the risks or costs associated with real-world testing. Simulations can also be scaled to test thousands of scenarios quickly, which is impractical in physical testing.

Who are Laplace Intelligence’s competitors?

Laplace’s most notable competitor is Applied Intuition, a company valued at $15 billion that has raised $1.2 billion to date. Other players in the simulation and testing space include Waabi, which focuses on autonomous vehicle development.

How does Laplace Intelligence make money?

Laplace generates revenue by providing simulation and testing services to customers in the autonomous vehicle and robotics industries. Its customers include companies like Isuzu, Bot Auto, and Nvidia, as well as partners in smart city initiatives.

What is the significance of Laplace’s smart city pilot in San Jose?

The smart city pilot in San Jose, conducted in collaboration with Nvidia and T-Mobile, demonstrates how Laplace’s simulation technology can be applied to urban planning and infrastructure. The project aims to showcase the potential of AI-driven simulations in managing traffic, congestion, and other smart city challenges.

Related on Lazyfounder

• OpenAI
• Google
• More technology news
• Arm’s AGI CPU Outperforms AMD in Agentic AI Benchmarks, Signal65 Reports