Photo: Magda Ehlers / Pexels
Safeworld exits stealth with $12M seed to stress-test robots around humans
The startup uses simulation and generative AI to find rare failure cases before robots meet real people
Safeworld is building digital humans so that robots stop endangering the real ones. The robotics safety startup is stepping out of stealth with a $12 million seed round behind it.
What Safeworld is building
Safeworld is developing a testing and evaluation platform for interactions between humans and robots. It uses simulation and generative AI to hunt for rare but critical failure cases, the moments when a machine does something dangerous that nobody thought to test.
Safeworld plans to pair its simulation work with real-world testing rather than replacing physical trials outright.
The company is aiming at enterprises deploying robots in environments where people are present. It also plans to work with the open-source and research communities on advancing safety standards for physical AI systems.
The money and the team
The $12 million seed round was completed around February 9, 2026. Backers include 515 Ventures, Umami Capital and Founders Future.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Safeworld also joined the a16z speedrun accelerator, which reportedly contributed $500K in April 2026.
The team numbers around six employees, based in the Oakland and Palo Alto area of California.
The team includes expertise from a Carnegie Mellon University professor focused on robot safety. That professor is also a recipient of the NSF CAREER Award, a National Science Foundation honor for early-career faculty.
The company was founded in either 2025 or 2026.