French robotics-focused start-up Genesis AI has officially released its first foundational model, GENE-26.5, along with a human-like robotic hand built in-house. The hand is capable of carrying out complex tasks such as cooking or solving a Rubik's Cube, marking a significant step forward in the field of embodied artificial intelligence. The launch represents a rare move to control both the AI software and the physical hardware, giving the company a unique advantage in the race toward general-purpose robots.
The company, which completed a $105 million (£77.2m) seed funding round in July of last year, began with a focus on building an AI model designed to power real-world robotics. According to co-founder and chief executive Zhou Xian, the decision to build its own hand came from a desire to have more accurate control over the entire stack, from the AI model to the physical hardware. 'If we only built software, we would be dependent on someone else's hardware,' Zhou said in an interview. 'By building our own hand, we can optimize the full pipeline of perception, cognition, and action.'
A hand that mirrors the human appendage
Most robotic hands developed by other companies tend to have two or three fingers, optimized for specific industrial grips. In contrast, the hand produced by Genesis AI closely resembles a human appendage, with five fingers and a fully articulated palm. This design choice is not merely aesthetic; it allows data collected from human demonstrations to be more directly transferred to the device, minimizing the gap between human motion and robotic execution.
The human-like form factor is expected to accelerate learning in tasks that require dexterity and fine motor skills. By mimicking the kinematics of a human hand, the robot can use the same movement strategies that people naturally employ, from twisting a jar lid to tying a knot. Genesis AI believes this biomimetic approach will be critical for achieving general-purpose manipulation in unstructured environments like kitchens, laboratories, and factories.
Each finger is driven by a set of miniature motors and tendon cables, allowing for natural compliance and backdrivability. The hand has 26 degrees of freedom, which is comparable to the complexity of a human hand. It is also equipped with tactile sensors distributed across the fingertips and palm, providing real-time feedback on pressure, texture, and slippage. This allows the robotic hand to adapt its grip dynamically, much like a human would when handling fragile objects or slippery surfaces.
GENE-26.5: a foundational model for robotics
At the heart of the system is GENE-26.5, a foundational model trained on vast amounts of data from both simulated and real-world demonstrations. The model is designed to interpret high-level instructions and break them down into low-level motor commands. It can also learn from video data, allowing it to acquire new skills by observing human workers or even online videos.
The model's architecture is based on a transformer network with billions of parameters, and it processes multimodal inputs including visual images, tactile signals, and proprioceptive joint encoders. This makes it capable of reasoning about spatial relationships, object properties, and the physical consequences of its actions. The model is also designed to support continuous learning, meaning it can improve its performance over time as it is exposed to more data.
One of the key challenges in robotics is the scarcity of high-quality training data. While many companies rely on simulation, Genesis AI argues that real-world data is irreplaceable. This is why the company has invested heavily in data collection infrastructure, including sensor gloves that can be worn by human workers to capture detailed hand movements. The company also uses teleoperation systems, where a human operator controls the robot from a distance, to generate demonstration data that can be used for imitation learning.
Sensor gloves and data collection
Genesis has also created lightweight sensor gloves that can be worn by workers in fields such as pharmaceuticals or manufacturing. These gloves capture fine-grained data on hand posture, finger forces, and movement patterns, which can then be passed along to the AI model for training. The gloves are equipped with inertial measurement units and flexible strain sensors that measure finger joint angles at high frequency, ensuring that the resulting dataset is rich and accurate.
The company says that data from the gloves can be combined with information gleaned from videos of workers carrying out tasks, providing a dense signal for imitation learning. This approach raises an interesting ethical question: are workers willing to help train a potential robotic replacement? The company has not yet addressed worker concerns directly, but it argues that the technology is intended to augment human capabilities rather than replace them entirely.
In the short term, robots could take over repetitive or ergonomically stressful tasks, allowing humans to focus on higher-level decision-making and quality control. For example, in pharmaceutical manufacturing, robots could handle hazardous chemicals while humans oversee the process. In agriculture, robots could harvest fragile fruits without bruising them, a task that has been notoriously difficult for traditional automation.
In a demonstration video, Genesis showed a pair of its hands carrying out complex or delicate processes such as preparing a smoothie or playing the piano. These demonstrations highlight the system's ability to handle both fine manipulation and dynamic tasks. The video also shows the hands recovering from errors in real time, such as adjusting a grip on a slippery object or repositioning the fingers to properly grasp a bottle.
Strategic leadership and European roots
Co-founder and president Theophile Gervet previously worked for France's Mistral, one of the leading European AI startups. His experience in the AI ecosystem has helped shape Genesis's approach to both research and commercialization. The company retains a substantial presence in Europe, which it attributes to the continent's large talent base and potential market of industrial customers.
European countries have been increasingly investing in robotics and artificial intelligence, with initiatives such as Horizon Europe providing funding for research and innovation. Genesis's dual presence in the US and Europe allows it to tap into the strengths of both regions: the rapid commercial adoption and venture capital environment of California, and the engineering talent and manufacturing expertise of Europe.
The team is young but highly accomplished. Zhou Xian was previously a research scientist at leading AI labs, and Gervet's track record in AI startups has helped the company attract top talent. The founding team includes experts in machine learning, mechanical engineering, and control systems, reflecting the interdisciplinary nature of the project.
Investor backing and growth
The seed funding round in July last year was co-led by Eclipse and Khosla Ventures, two prominent venture capital firms with strong track records in deep tech and AI. Additional backing came from Bpifrance, HSG, former Google chief Eric Schmidt, telecoms entrepreneur Xavier Niel, MIT Computer Science and Artificial Intelligence Laboratory director Daniela Rus, and Apple distinguished scientist Vladlen Koltun.
This constellation of investors includes both corporate figures and academic luminaries, signaling confidence in Genesis's technical vision. The $105 million round is one of the largest seed rounds on record for a robotics company, reflecting the intense interest in AI-enabled robotics. The funding is being used to expand the engineering team, accelerate research, and develop the next generation of hardware.
Genesis's team of 60 people is split roughly evenly between the US and Europe, with a slightly larger group based in San Carlos, California. The company also has offices in Paris and London. The San Carlos location places it in the heart of California's robotics ecosystem, alongside companies like Figure, 1X, and Tesla's Optimus. The European offices allow close collaboration with industrial clients and research institutions.
Traction in an increasingly crowded field
The market for humanoid robots and dexterous manipulation is becoming increasingly crowded. Companies like Tesla, Boston Dynamics, and a host of startups are all pursuing the goal of a general-purpose robot. However, most of these companies have focused on full-body humanoids with simpler grippers or two-fingered end effectors. Genesis's decision to prioritize the hand sets it apart.
'The hand is the hardest part,' says Zhou. 'Once you can achieve human-level dexterity, the rest of the robot becomes much easier to manage.' The company plans to integrate the hand into a full-body robot capable of walking, navigating, and performing tasks across different environments. While the company has not provided a timeline, it confirmed that the full-body robot is a strategic priority.
The GENE-26.5 model is also designed to be hardware-agnostic, meaning it can potentially be licensed to other robot manufacturers. This would create a new revenue stream for the startup and could accelerate the adoption of its foundational model across the industry. However, competitive concerns may arise if other companies see Genesis as both a hardware partner and a future rival.
Regulatory and societal considerations
As with any disruptive technology, the development of human-like robotic hands raises important regulatory and societal questions. The European Union has been working on comprehensive AI regulation that will address both safety and ethical aspects. Genesis's transparent approach to data collection and its emphasis on safety features could help it navigate these regulations more easily.
There are also concerns about data privacy and consent when deploying sensor gloves in industrial settings. Workers must be clearly informed about what data is being collected and how it will be used. Genesis says it is committed to ethical data practices, and it has an advisory council that includes academics and ethicists.
The company is also considering open-sourcing its training datasets and some of its model architectures, similar to how Meta has released open models like LLaMA. This could foster innovation and attract a community of developers, but it also carries risks related to misuse and unintended consequences. A measured approach is likely, with the company planning to release a subset of its data first.
Looking ahead
Genesis AI's debut marks a major milestone in the field of AI robotics. By combining cutting-edge foundational models with a highly dexterous, human-like hand, the company has demonstrated that it is possible to create a system that can learn and perform a wide range of tasks. The path from demonstration to commercial deployment is still long, but the progress so far is impressive.
The company will need to scale its engineering team, expand its data collection efforts, and refine the manufacturing process for its robotic hand. It will also need to navigate a complex competitive landscape and prove that its technology can deliver real-world value in industrial applications. With strong financial backing and a senior team that combines research excellence with startup agility, Genesis AI is well positioned to become a key player in the next wave of robotics innovation.
As it continues to develop its full-body platform and improve its AI model, the industry will be watching closely. The ability to replicate the intricate functions of the human hand could unlock new possibilities in automation, healthcare, and beyond. For now, the company is focused on perfecting the technology and building a foundation that will define the future of embodied AI.
Source: Silicon UK News