Safeworld, a startup focused on ensuring the safety of generative AI-powered robots, emerged from stealth on Monday with a seed funding round exceeding $12 million. The investment was led by Shine Capital and Andreessen Horowitz Speedrun, with additional contributions from Box Group, the Carnegie Mellon University Endowment, Innovation Endeavors, and SV Angel.
The company addresses a critical bottleneck in robotics: while generative AI is increasingly used to control humanoid machines, its probabilistic nature makes it difficult to guarantee predictability and safety. Dr. Ding Zhao, who directs the Safe AI Lab at Carnegie Mellon University, co-founded Safeworld alongside veteran startup executive Kyle Wong and machine learning engineer Simo Rachidi.
“The safety challenge is a combination of two things,” Zhao explained. “First, you need really advanced generative AI probabilistic evaluations to underwrite the risk of a probabilistic system. The second part is the trust factor, and you need both to deploy a robot.”
Safeworld specializes in evaluating robotic control systems within simulations populated by realistic human models. This process mirrors the challenges faced by autonomous vehicle companies like Tesla and Wayve, which must ensure their vehicles react appropriately to unexpected road incidents. However, Zhao argues that testing robots is more complex because they operate in unstructured environments where safety standards vary by facility.
Wong cited the example of a blind corner in a factory. “If a human is carrying boxes, will the robot detect them? What speed or stopping distance is required to prevent a collision?” he asked. To answer such questions, Safeworld creates digital replicas of these environments using physics engines like Genesis or MuJoCo, inserts the robot driven by its actual software, and runs thousands of scenarios involving human models. These simulations also cover unpredictable events, such as trips and falls, which would be dangerous and impractical to test in reality.
Jonathan Lai, a partner at a16z Speedrun, emphasized the urgency of establishing industry standards early. “The time to build an industry safety standard is now while robots are being designed and deployed,” Lai told TechCrunch. “By the time you have robots in households colliding with kids and causing safety incidents, that’s way too late.”
Zhao noted that while some robot manufacturers build internal validation tools, many underestimate the difficulty of solving edge cases. “It is not the robot in the vacuum, in the demo, that we are worried about. It is the robot that is deployed at scale, with people who potentially never operated a robot before,” he said.
Gritt Robotics, a company developing AI brains for robots that install photovoltaic panels at solar farms, has partnered with Safeworld. Gritt CTO Vishal Dugar highlighted the limitations of traditional mathematical verification for such systems. “It necessarily has to be done empirically,” Dugar said, pointing out the need to account for diverse human appearances, postures, and behaviors like running or crouching on construction sites.
The startup is currently determining whether to offer its technology as a self-serve platform or a services-based model. Despite these early decisions, Zhao expressed confidence in the company’s trajectory. “We’ll probably be the first profitable company in this field,” he said. “Because if anyone wants to deploy, they need to pay us to handle the situation.”
Carnegie Mellon backing gives this serious credibility. The industry needs standards before incidents happen.
Wait, will this cost extra for every robot deployment? That could get expensive quickly for small manufacturers.
Simulation is the only way to handle unstructured environments safely. Smart bet on the safety bottleneck.