Safeworld raises $12M seed to validate generative AI robot safety

A new startup called Safeworld wants to solve a tricky question: how do you certify that a robot powered by generative AI is safe? The company emerged from stealth on Wednesday with a seed round of more than $12 million, TechCrunch reports.

Safeworld was founded by Dr. Ding Zhao, who runs the Safe AI lab at Carnegie Mellon University, along with startup veteran Kyle Wong and machine learning engineer Simo Rachidi. 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 core problem, Zhao told TechCrunch, is that generative AI systems aren't predictable the way traditional algorithms are. You need both probabilistic safety evaluations and human trust to deploy a robot at scale.

Safeworld's approach is to run simulations of robotic control systems in environments filled with realistic human models. The company builds a digital version of a factory corner or home hallway, inserts the robot's real software, and runs thousands of scenarios where simulated humans trip, crouch, carry boxes or run.

"Tripping and falling is a good example of something that we do a lot of testing with the simulation," Wong said. "Otherwise, you would have to go and trip and fall for the robot, which is like a hard thing to be doing all the time."

The challenge is harder than autonomous driving, Zhao argued, because robots work in unstructured spaces and each facility has different safety standards. a16z Speedrun partner Jonathan Lai told TechCrunch that now is the time to build safety standards, before robots are in homes colliding with kids.

Robot builders already run internal safety tests, but Safeworld's founders believe third-party validation will become necessary. Vishal Dugar, CTO of Gritt Robotics, is partnering with Safeworld. His company builds AI brains for solar-farm robots that work alongside human workers. Proving safety empirically, he said, requires considering all possible human behaviors and appearances.

Safeworld is still deciding between a platform or a services-based business model. Zhao is confident: "We'll probably be the first profitable company in this field, because if anyone wants to deploy, they need to pay us to handle the situation."

• Ding Zhao
• generative AI
• robot safety
• Safeworld
• seed funding
• simulation