Unbiased AI-powered news
Genesis AI has released GENE-26.5, a system that combines a robotics foundation model, a human-scale dexterous hand and a new data engine. The company said the technology enables general-purpose robots to perform complex physical tasks with human-level manipulation after learning from human movement data. Early deployments are expected in industrial settings later this year.
Fox NewsGenesis AI has unveiled GENE-26.5, a robotic brain it says enables general-purpose robots to perform complex physical tasks with human-level dexterity. The company, which describes itself as a global full-stack robotics firm, said the system pairs a robotics foundation model with a human-scale dexterous robotic hand and includes a new data engine.
Together the components allow robots to learn from human movement and handle tasks requiring precision and coordination. Theo Gervet, Co-Founder and President of Genesis AI, said the easiest way to understand the system is to view it as the component guiding the robot's actions.
"Think of GENE-26.5 like a robotic brain that takes in information and tells the robot what to do," Gervet said. "It is the industry's most advanced robotic brain, with the most advanced capabilities. The company developed a method to feed GENE-26.5 large volumes of data on human hand movements so the system can direct robotic hands to replicate them.
"GENE-26.5 can also tell our robotic hands how to do tasks with many, many steps," Gervet said. He cited an example in which a robot powered by the system follows a 20-step process to prepare a full omelet from start to finish. The company built a robotic hand that matches the form and function of a human hand.
It pairs with a glove that captures motion and pressure data. Genesis AI said the glove is 100 times cheaper than typical options and has shown up to five times greater data collection efficiency compared with traditional methods. "When a human wears the gloves as they interact with objects or do their work, we can capture details about the exact movements their fingers and wrists make," Gervet said.
The company said its simulation platform accelerates development by allowing the AI to train in a virtual environment before real-world testing. Gervet said robots have historically faced a data problem because training information from human hands did not transfer well unless the robotic hand closely matched human anatomy.
The company addressed that by designing a robotic hand that exactly matches a human hand.
AI expects the first applications to be in warehouses and manufacturing facilities. Gervet said the company is already in discussions with industrial customers and plans a phased rollout that later includes service industry and consumer home use. "In addition, we’re hoping that in a home setting, our technology will be able to help handle daily chores, freeing up time for people to spend doing what they actually enjoy," Gervet said.
The company said safety testing is central to development. Systems undergo extensive validation first in simulation running millions of scenarios and then in controlled real-world environments. Genesis AI said it follows established safety standards and industry regulations for robots operating around people.
The company is currently showcasing individual components including the robotic brain, robotic hands and data collection system. It plans to unveil a full general-purpose robot that integrates all elements. Early, small-scale deployments with select partners could begin later this year.
Single source — no framing comparison available.
moneycontrol.comSony Music Entertainment sued AI music generator Udio in New York court on Monday. The complaint targets unauthorized use of more than 30,000 recordings to train the company's models.
sbs.com.auFrench lawmakers reached a compromise Monday on a bill that would bar children under 15 from social media accounts. The measure is expected to pass Tuesday and take effect in stages starting September.
newatlas.comZ.ai's GLM-5.2, released in June 2025, ranks as the top open-source model by some benchmarks and fifth on OpenRouter usage. DeepSeek's earlier R1 release in January 2025 triggered a trillion-dollar drop in U.S. tech values and calls for bans.