Safeworld Launches to Tackle Safety Risks of Generative AI‑Powered Robots

Safeworld was founded by Dr. Ding Zhao of Carnegie Mellon’s Safe AI Lab, veteran executive Kyle Wong and ML engineer Simo Rachidi to address the unpredictability that comes when generative AI models control robotic systems.

The company uses high‑fidelity simulators such as Genesis or MuJoCo to create digital twins of real‑world workspaces, inserts the robot’s actual control software, and runs thousands of scenarios with varied human models—including edge cases like tripping, blind corners, or obscured limbs—to evaluate collision risk and stopping distances.

Founders argue that a third‑party safety validator is essential for sharing safety cases across competitors and for establishing an industry‑wide standard while robots are still in the design phase.

Why it matters for GPU / AI infrastructure

Running massive numbers of physics‑based simulations demands GPU‑accelerated workloads; robust safety testing will drive demand for high‑performance AI hardware that can handle complex, probabilistic robotics workloads at scale.