Safeworld has emerged from stealth with more than $12 million in seed funding to test AI-controlled robots against simulated people and workplaces. The round was led by Shine Capital and a16z Speedrun, with participation from several other investors.
The company builds a digital version of a deployment site, inserts a simulated robot running its real control software and tests thousands of encounters. A factory scenario might vary a person’s position, clothing, movement or behavior near a blind corner to measure whether the robot detects them and stops in time. Falls and other risky situations can be repeated without asking people to perform them physically.
Carnegie Mellon Safe AI Lab director Ding Zhao founded the company with Kyle Wong and Simo Rachidi. Solar-construction robotics company Gritt Robotics is working with Safeworld while developing machines that operate alongside installers.
Simulation is useful because generative control systems are probabilistic and many edge cases are impractical to reproduce safely. It is not proof that a robot will behave identically in the physical world: sensors, surfaces, lighting and unexpected human actions can differ from the model. Safeworld is still deciding whether to operate mainly as a software platform or a service, and independent testing will matter most if its scenarios and failure criteria are transparent.