Assured Robot Intelligence (ARI) is a humanoid-robotics startup Meta acquired to strengthen AI models for robots and humanoids. It built foundation AI models, large models trained to control varied tasks, for humanoids doing general physical labor in busy human spaces. TechCrunch reported that Meta completed the deal on May 1, 2026, with terms undisclosed, in the TechCrunch report on the acquisition.
The full ARI team joined Meta Superintelligence Labs and works with Meta Robotics Studio. That move places ARI's researchers inside Meta's humanoid-technology group formed in 2025. Mobile World Live described that team transfer in the Mobile World Live account of the deal.
Table of Contents
- What did ARI build?
- Who started ARI and where are they now?
- How does ARI fit Meta's humanoid plan?
- What should robotics readers watch next?
What did ARI build?
ARI focused on robot brains rather than robot bodies. Its work centered on foundation models for humanoids performing household chores and other physical work. The goal was operation in complex rooms occupied by people. Meta described ARI as working at the frontier of robotic intelligence for understanding, predicting, and adapting to human behavior.
Humanoid Guide quoted that description in its summary of Meta's statement. That skill matters because homes, clinics, and warehouses change constantly. The team's assigned work at Meta covers robot control, self-learning models, and whole-body humanoid movement. Whole-body movement means coordinated legs, torso, and arms during walking, carrying, and bending. For readers, ARI is best understood as a controls-and-learning bet.
Who started ARI and where are they now?
ARI's co-founders are Lerrel Pinto and Xiaolong Wang. Pinto was co-founder and CEO of ARI and now leads frontier robotics models at Meta Superintelligence Labs. He also runs NYU's General Robotics and AI Lab and previously co-founded Fauna Robotics. Wang brought research experience from Nvidia and academic work on robot learning and computer vision.
He served as an associate professor at UC San Diego, while Pinto taught at NYU. That pairing joined academic robot learning with industry vision work. Both founders moved to Meta with the acquisition. Their group collaborates closely with Meta Robotics Studio. Readers following humanoid talent should track Pinto's lab and Meta's Superintelligence Labs output.
How does ARI fit Meta's humanoid plan?
Meta is pursuing an Android-like platform strategy for humanoids. It is developing in-house hardware, sensors, and licensable software. The practical aim is software other companies can use on varied humanoid bodies. That approach responds to a software bottleneck.
Meta CTO Andrew Bosworth framed software as the bottleneck in 2025. WinBuzzer covered that platform framing in the WinBuzzer report on Meta's humanoid plans. ARI supplies model and control expertise for that stack. Its background in high-value labor markets points toward chores, logistics support, and repetitive physical tasks. Expect early value in improved perception, prediction, and safer motion around people.
What should robotics readers watch next?
Treat this deal as research and talent first. Meta kept price, hardware specifications, product dates, and deployment plans out of its announcement. No buyer should plan purchases around an ARI-branded robot.
Automation teams can still act now: ARI had raised a seed round from AIX Ventures before the sale. That early-stage backing fits an acqui-hire pattern focused on people and models. The useful signal is direction, not a shipping product.
- Follow Meta Robotics Studio for software interfaces and sensor choices
- Test whether foundation-model control reduces task-specific programming
- Measure human-aware navigation in crowded test cells before pilot rollout
- Separate demo clips from repeatable shifts with people nearby
