Figure AI, the Silicon Valley-based robotics company perhaps best known for its collaboration with BMW, has laid out a structured approach to improving safety for humanoid robots operating in workplace environments. The plan, as detailed in the company's public communications, is built around the premise that advanced AI-driven autonomy — rather than purely mechanical safeguards — will be the primary lever for making bipedal machines safe to work alongside human employees.
The core of the initiative rests on several pillars. First, Figure AI is focusing on the continuous optimization of robot-generated data. This means that every shift a humanoid works, every movement it makes, and every interaction it has with its environment produces data that can be fed back into the system to improve future performance. This is not a one-time calibration exercise but an ongoing loop of collection, analysis, and refinement.
Second, the company plans to share learned behaviors across its robot fleet. If one unit at a particular facility discovers a more efficient or safer way to handle a task, that knowledge can be propagated to other units, potentially at different sites. This fleet-wide learning approach is designed to accelerate the pace at which safety improvements are realized, moving beyond the limitations of single-unit programming.
Third, Figure AI is exploring adjacent applications as its AI capabilities expand. The idea is that the underlying technology powering the humanoid — the perception systems, the decision-making algorithms, the motion planning — can be adapted to new tasks and new environments without requiring a complete redesign. This flexibility is intended to make the robots more versatile and, crucially, more adaptable to the specific safety requirements of different workplaces.
The announcement comes at a time when humanoid robots are transitioning from laboratory curiosities to commercially deployed assets. The Figure 02, the second iteration of the company's humanoid, has already been trialed in real-world production settings at BMW's Spartanburg plant in South Carolina. That deployment, which began in earnest around 2025, has been described as among the first major automaker trials of humanoid robots in active production environments. The Munich-based manufacturer has been testing these robots to enhance flexibility, address labor shortages, and automate tasks that were previously considered beyond the reach of traditional automation.
The broader context is that the humanoid robot market is becoming increasingly crowded. Figure AI's Figure 02 is priced in the $50,000 to $80,000 range and is positioned for manufacturing and assembly work, with its Helix AI system and the BMW partnership cited as key strengths. Tesla's Optimus Gen 2, with a target price of $20,000 to $30,000, is aimed at general workplace tasks and is currently in a pilot phase. Agility Robotics' Digit, priced between $100,000 and $150,000, is purpose-built for logistics and warehousing, specifically tote handling, and is already commercially available. Each of these machines represents a different approach to the same fundamental question: how do you build a machine that can navigate human-built environments, use existing tools, and adapt to dynamic work settings without requiring infrastructure modifications?
Figure AI's safety plan is, in part, a response to that question. By emphasizing AI-driven autonomy, the company is positioning safety not as a static feature but as a dynamic property of the system. The robots are designed to learn from their environments, to adapt to changing conditions, and to improve over time. This is a significant departure from traditional industrial robots, which typically operate in caged-off areas with rigid, pre-programmed motion paths.
Why it matters for European robot service
For the European robotics ecosystem, Figure AI's safety initiative carries implications that extend well beyond the company's own product line. The European market has been a proving ground for industrial automation, with automotive manufacturing in Germany, logistics operations in the Netherlands, and a broad range of service applications across the continent. The adoption of humanoid robots in these settings is not hypothetical; it is already underway, and the safety frameworks being developed now will shape how these machines are deployed, certified, and maintained.
The automotive sector is particularly instructive. BMW's collaboration with Figure AI is a concrete example of a European manufacturer integrating humanoids into active production. The company has been testing these robots in real-world settings to improve flexibility and address labor shortages. The expectation, according to the source material, is that automotive production will very quickly come to see humanoids as an integral part of their manufacturing processes. This is not a distant prospect; it is a near-term trajectory.
The source material also notes that full-bodied humanoid robots are likely to be employed due to their mobility advantages, allowing them to quickly take over the role of traditional AGVs (automated guided vehicles) and AMRs (autonomous mobile robots) in automotive production environments. This is a significant shift. AGVs and AMRs have been workhorses of factory automation for decades, moving materials along fixed or semi-fixed paths. Humanoids, by contrast, can navigate stairs, open doors, and use tools designed for humans. They can, in principle, step in where wheeled platforms cannot go.
But this mobility advantage comes with a safety burden. A robot that can move freely through a human workspace is a robot that can, in theory, collide with a human. The source material is explicit on this point: as robots increasingly operate alongside humans in factories and service settings, ensuring they operate safely is not just important, it is essential for the robotics industry. The AI-driven autonomy fundamentally changes the safety landscape, making testing, validation, and human oversight much more complex — but also more necessary.
For European robot service providers, this creates both challenges and opportunities. The challenges are technical: how do you validate the safety of a system that is continuously learning and adapting? How do you certify a machine whose behavior is not fully predetermined? The source material notes that robotic systems need to be designed and certified in line with ISO safety standards, but the application of those standards to AI-driven humanoids is an open question. The opportunities are commercial: as more humanoids are deployed, there will be a growing need for service, maintenance, and safety auditing. The companies that can develop expertise in these areas will be well-positioned.
There is also a workforce dimension. The source material observes that AI in robotics will further influence how teams work, how decisions are made, and how performance is monitored. This can improve workflows but may also raise concerns about employee surveillance or reduced autonomy. Companies and governments are pushing reskilling and upskilling programs to help workers adapt. For European service providers, this suggests a growing market for training and consulting services, as organizations seek to integrate humanoids without alienating their human workforce.
The source material also highlights a notable commercial development: in 2025-01, Brett Adcock, founder of Figure AI, announced that the company had signed its second commercial customer, described as "one of the biggest US companies." The name of that customer was not disclosed in the source material. What is known is that Figure AI's first commercial customer was BMW, and the second is a major US firm. This expansion of the customer base is a signal that humanoid robots are moving beyond pilot programs and into broader commercial deployment.
What buyers and operators should know
For organizations considering the adoption of humanoid robots, the source material provides a framework for understanding both the potential and the limitations of current technology. The table of available models is a useful starting point, but it is important to read it carefully. The price ranges are not fixed; they reflect the current state of the market and, in the case of Tesla, a target rather than an actual price. The "availability" column is equally telling: Figure 02 is in limited deployment, Tesla Optimus Gen 2 is in a pilot phase, and Agility Digit is commercially available. These distinctions matter for planning purposes.
The source material outlines a four-phase approach to deployment, with the final phase being ongoing optimization. This phase involves leveraging robot-generated data for continuous improvement, sharing learned behaviors across the robot fleet, exploring adjacent applications as AI capabilities expand, and planning for next-generation upgrades. For buyers, this implies that the robot you purchase today is not the robot you will be operating a year from now. The software will evolve, the behaviors will improve, and the range of tasks the robot can perform will expand. This is a double-edged sword: it means the robot can become more valuable over time, but it also means that the operator must be prepared for a continuous process of updates, retraining, and revalidation.
The source material is also candid about the challenges and limitations that workplace humanoid robots face in 2026. Despite rapid progress, these machines are not yet plug-and-play solutions. Organizations must plan for the realities of deployment, which include the need for testing, validation, and human oversight. The source material notes that AI-driven autonomy makes these processes more complex, not less. This is a critical point for operators to understand: a humanoid robot is not a traditional industrial robot with a safety cage. It is a system that must be integrated into the human workflow, and that integration requires careful planning.
One of the key technical developments noted in the source material is the role of generative AI in how humanoids acquire capabilities. The robots can learn from demonstration and even figure out tasks independently. This is a transformative shift in how robots are programmed. Traditional industrial robots require explicit programming for each task; humanoids with generative AI can, in principle, observe a task being performed and then replicate it. This could transform the way traditional robots are programmed and pave the way for new application scenarios. For operators, this means that the barrier to entry for new tasks is lower, but it also means that the robot's behavior is less predictable, which has safety implications.
The source material does not disclose specific safety metrics, incident rates, or certification details. It does not provide SLA numbers, response times, or spare-part lead times. These are important gaps. Buyers should be aware that the source material describes the company's plan and the general state of the industry, but it does not provide the kind of granular data that would be needed for a formal procurement decision. Organizations considering humanoid robots should seek additional information from the manufacturers, including detailed safety documentation, test results, and references from existing customers.
Another consideration is the economic case. The source material notes that humanoid robots are likely to be employed in automotive production due to their mobility advantages, allowing them to take over the role of traditional AGVs and AMRs. This suggests that the business case is not necessarily about replacing human workers but about replacing or augmenting existing automation. A humanoid that can do the work of an AGV, but with greater flexibility, may offer a better return on investment in certain settings. However, the price points are significant: even at the lower end, a humanoid robot represents a substantial capital investment. The total cost of ownership, including maintenance, software updates, and potential downtime, must be factored into any decision.
The source material also touches on the broader societal implications. As AI in robotics influences how teams work, how decisions are made, and how performance is monitored, there are legitimate concerns about employee surveillance and reduced autonomy. Companies and governments are pushing reskilling and upskilling programs to help workers adapt. For operators, this means that the introduction of humanoid robots is not purely a technical decision; it is also a human resources decision. The workforce must be prepared for the change, and the organization must have a plan for how humans and robots will interact on a daily basis.
In summary, the source material presents a picture of an industry in transition. Humanoid robots are moving from the lab to the factory floor, and safety is emerging as the central challenge. Figure AI's plan to use AI-driven autonomy as the primary safety mechanism is a significant statement of intent, but it is not a complete answer. The industry as a whole is still grappling with how to test, validate, and certify these systems. For European buyers and operators, the message is clear: the technology is promising, the momentum is real, but the due diligence requirements are substantial. The source material provides a solid foundation for understanding the landscape, but it is not a substitute for detailed, case-by-case evaluation.
Sources
https://www.bundle.app/en/technology/figure-ai-details-plan-to-improve-humanoid-robot-safety-in-the-workplace-90af3455-30d1-4dd9-9c73-060f447e0242
Published by Vigla Media OÜ (Estonia).