Safeworld Launches With $12M to Test AI Robots in Safety Simulations Before They Reach Humans
Safeworld runs robots’ actual software through thousands of simulated scenarios featuring realistic human behavior, from blind factory corners to falls. Founded by Carnegie Mellon’s Ding Zhao, it raised more than $12 million from Shine Capital and a16z Speedrun.
Safeworld, a startup focused on evaluating the safety of generative AI-powered robots, emerged from stealth in October 2026 with a seed round of more than $12 million. The round was led by Shine Capital and a16z Speedrun, with participation from Box Group, Carnegie Mellon University Endowment, Innovation Endeavors, and SV Angel.
The company was founded by Dr. Ding Zhao, who directs the Safe AI Lab at Carnegie Mellon University, alongside veteran startup executive Kyle Wong and machine learning engineer Simo Rachidi. Safeworld’s core offering is simulation-based safety testing: it builds digital environments — including realistic human models — and runs thousands of scenarios to evaluate how a robot’s control system responds to unpredictable human behavior.
The approach addresses a specific problem with generative AI-driven robots. Unlike traditional algorithms, generative AI systems are probabilistic, making their behavior harder to predict. “It is not the robot in the vacuum, in the demo, that we are worried about,” Zhao said. “It is the robot that is deployed at scale, with people who potentially never operated a robot before.”
Scenarios tested in simulation include robots navigating blind corners in factories, detecting humans carrying boxes, and responding to people who trip or fall. Safeworld uses physics simulation platforms such as Genesis and MuJoCo and runs the robot’s actual software inside those environments. The company argues that robot manufacturers will also want independent third-party validation, not just internal testing tools.
Gritt Robotics, which develops AI systems for robots used at industrial-scale solar farms, is already partnering with Safeworld. Gritt’s CTO, Vishal Dugar, noted that formally proving safety through mathematical verification alone is insufficient for these systems and that empirical testing is necessary.
Safeworld is still determining its business model — whether to operate as a platform or a services company — but Zhao suggested the need is clear: any company seeking to deploy robots will require safety validation before doing so.