French robotics-focused startup Genesis AI has taken a significant step toward general-purpose robotics by releasing its first foundational artificial intelligence model, named GENE-26.5, alongside an intricately human-like robotic hand built entirely in-house. The hand, which closely mimics the dexterity and size of a human appendage, is capable of performing complex and delicate tasks such as cooking a meal, playing the piano, or solving a Rubik's Cube. This development marks a departure from traditional robotic grippers, which typically use only two or three fingers, and opens up new possibilities for precise manipulation in industrial and domestic settings.
The company, which completed a $105 million (£77.2 million) seed funding round in July of last year, began with a primary focus on creating an AI model to power real-world robots. Co-founder and chief executive Zhou Xian explained that the decision to build the robotic hand from scratch was driven by the need for tighter integration and control over the entire hardware-software stack. “If you rely on third-party hardware, you are limited by someone else's design. We needed full stack ownership to achieve the level of precision and performance our model demands,” Zhou stated during a product briefing.
Human-like Design for Better Data Transfer
Unlike robotic hands developed by dozens of other companies, which tend to have two or three fingers and limited range of motion, the unit from Genesis AI closely resembles a human hand in size, joint structure, and sensory capability. This design allows data collected from human demonstrations—such as via sensor gloves or video recordings—to be more directly transferred to the robotic system. The hand includes multiple degrees of freedom per finger, tactile sensors, and force feedback mechanisms that enable it to handle fragile objects without damaging them.
In a demonstration video released by the company, a pair of Genesis AI hands is shown working in coordination to prepare a smoothie—selecting fruits, operating a blender, and pouring the drink into a glass. Another clip features the hands playing a simple melody on a piano. The company also showcased the hand solving a Rubik's Cube in under a minute, a task that requires precise rotations and adjustments. These demonstrations highlight the potential of the GENE-26.5 model to learn and generalize manipulation tasks from limited datasets.
Sensor Gloves for Data Collection
To accelerate the training of its AI model, Genesis AI has also created lightweight sensor gloves that can be worn by workers in fields such as pharmaceuticals, manufacturing, or logistics. These gloves capture detailed motion data—finger movements, grip forces, and hand orientations—which is then passed to the AI model for learning. The company emphasized that combining glove data with information gleaned from standard video recordings of workers performing tasks provides a rich multimodal dataset that helps the model understand complex sequences.
However, the introduction of such a data collection method raises questions about worker privacy and the potential for job displacement. When asked whether employees might be reluctant to help train a technology that could eventually replace them, Genesis AI co-founder and president Theophile Gervet acknowledged the concern but argued that the technology is intended to augment human capabilities first. “We see the gloves as a tool for improving worker ergonomics and safety in the short term. The data helps us build robots that can take over dangerous or repetitive tasks, but the path to full automation is long and will involve collaboration with humans,” he said.
Background and Team
Genesis AI’s roots in Europe remain strong, with the company maintaining substantial operations in France and the United Kingdom. Co-founder Theophile Gervet previously worked for Mistral AI, another French artificial intelligence startup, and the team draws heavily from Europe’s deep talent pool in robotics and machine learning. The company’s 60-person team is split roughly evenly between the United States (headquartered in San Carlos, California) and Europe (offices in Paris and London). The European presence is attributed to the continent’s strong research universities, open innovation ecosystems, and a large market of industrial customers in manufacturing, pharmaceuticals, and logistics.
The company’s focus on hardware alongside AI is relatively unusual among robotics startups, which often choose to license hardware and focus purely on software. Gervet explained that building the hand in-house allowed the team to overcome limitations in off-the-shelf components and achieve better performance. “By controlling the design of every muscle, sensor, and actuator, we can push the boundaries of what the model can learn,” he said.
Full-Body Robot in Development
Genesis AI has also revealed that it is working to supplement the mechanical hand with a full-body, general-purpose robot. While no timeline was given for the complete humanoid, the company’s roadmap suggests a multi-year development process. The goal is to create a robot that can navigate real-world environments, perform manipulation tasks, and interact safely with humans. The GENE-26.5 model is designed to serve as the brain for such a robot, integrating vision, language, and motor control functions.
The development of humanoid robots has gained momentum globally, with companies such as Tesla, Figure AI, and Boston Dynamics also pursuing similar goals. Expert opinions on the feasibility of general-purpose humanoids vary widely, but Genesis AI’s approach of building a highly dexterous hand first—rather than starting with a full body—is seen as a practical step by many roboticists. “Dexterous manipulation is one of the hardest problems in robotics. If they can crack that, integrating a mobile base and arms becomes more straightforward,” said Professor Maria Gomez, a robotics researcher at the Technical University of Munich who is not affiliated with Genesis.
Funding and Investors
The $105 million seed funding round last year was co-led by Eclipse and Khosla Ventures, with additional backing from a notable roster of investors: Bpifrance (the French public investment bank), HSG (Hasso Plattner Institute fund), former Google CEO Eric Schmidt, telecom entrepreneur Xavier Niel, MIT Computer Science and Artificial Intelligence Laboratory director Daniela Rus, and Apple distinguished scientist Vladlen Koltun. This mix of institutional, corporate, and individual investors reflects the high-stakes nature of the robotics AI market. The company has not announced any subsequent funding rounds, but industry analysts estimate that the capital will be sufficient for 2-3 years of development given the team size.
The involvement of Eric Schmidt and Daniela Rus signals strong interest from top-tier AI experts. Schmidt has been an active investor in artificial intelligence companies since leaving Google, while Rus is a leading authority on robot learning and manipulation. Koltun, who previously worked on reinforcement learning at Apple, brings expertise in training models for complex tasks. Genesis AI’s ability to attract such talent is indicative of the potential seen in its approach.
Market Implications and Challenges
If successful, Genesis AI’s humanoid robot could address labor shortages in manufacturing, healthcare, and logistics. The ability to perform fine motor tasks like handling syringes, assembling small components, or sorting delicate items would make it valuable in sectors that currently lack automation. However, the company faces significant competition from other startups and tech giants, as well as technical hurdles related to battery life, cost, and safety certification. Moreover, the ethical questions surrounding data collection via worker-worn gloves may require regulatory navigation. The company states that it is committed to transparent data handling and will work with labor representatives.
The robotic hand itself is not yet available for commercial sale, but Genesis AI plans to offer it as a development platform for researchers and industrial partners later this year. The GENE-26.5 model, meanwhile, is available under a research license to academic institutions, and the company is exploring a cloud API for commercial evaluation. As the field of embodied AI continues to heat up, Genesis AI’s move to release both a capable model and a purpose-built hand positions it as a serious contender in the race to build the first practical general-purpose robot.
Source:Silicon UK News