Safety Safety
A humanoid robot is heavy, strong and moves freely through space. Unlike an industrial robot behind a safety fence, it shares the workplace with people – and it can fall over. Safety is therefore the precondition for every deployment.
Key topics
- Collaboration
- Power and force limiting, speed and separation monitoring and compliant drives reduce the risk of injury on contact.
- Stability & falls
- Bipeds are dynamically stable: if control fails, the robot tips over. Fall detection, safe lowering and protective zones are key topics.
- Standards
- Relevant standards include ISO 10218 (industrial robots), ISO/TS 15066 (collaboration) and ISO 13482 (personal care robots). A dedicated standard for dynamically stable, legged robots is being developed (ISO 25785-1).
- Functional safety
- Safety functions such as emergency stop or safety-rated monitored standstill must work even when AI components behave unpredictably.
Key questions
- Which residual risks remain – and who accepts them?
- How do you certify a learning system?
- How does the robot behave during a power failure or a fall?
Governance & Legal Law & regulation
Humanoid robots fall under several sets of rules at once: as machinery, as an AI system, as a product with software and as a data processor with cameras and microphones. Anyone developing or deploying them needs a clear governance model.
Key topics
- EU AI Act
- A risk-based framework for AI. With the “AI Omnibus” adopted in 2026, obligations for high-risk systems were postponed: stand-alone systems from December 2027, AI embedded in products from August 2028.
- Machinery Regulation
- The EU Machinery Regulation (EU) 2023/1230 applies from 20 January 2027 and, for the first time, explicitly addresses machines with self-evolving behaviour.
- Product liability
- The new Product Liability Directive (EU) 2024/2853 explicitly covers software and AI – even updates after sale can be relevant for liability.
- Data protection & cyber
- Cameras and microphones process personal data (GDPR). Connected robots are also subject to cybersecurity requirements (Cyber Resilience Act).
Key questions
- Who is manufacturer, operator, provider – and who is liable?
- What documentation does conformity assessment require?
- How do we keep track of deadlines and transition rules?
Overview only, not legal advice. Deadlines and details change – always check the current status.
Ethics Ethics
The more human a machine appears, the more strongly we respond to it emotionally. That creates opportunities – in care and assistance, for example – but also risks of deception, dependency and surveillance.
Key topics
- Honesty, not deception
- Robots should remain recognisable as machines. A human-like appearance must not suggest abilities or feelings that do not exist.
- Work & participation
- Humanoids take on monotonous, heavy or dangerous tasks. What matters is how companies and society shape the change in jobs.
- Privacy
- A robot with cameras in a living room or on a hospital ward sees and hears a lot. Data minimisation and transparency are basic conditions for acceptance.
- Responsibility
- When a learning system makes mistakes, there must be no responsibility gap: roles, oversight and the ability to intervene must be clear.
Key questions
- Where should humanoid robots deliberately not be used?
- How much human likeness makes sense?
- Who benefits – and who bears the costs?
Technology Technology
The current momentum comes from combining powerful hardware with AI models that link language, images and movement (“physical AI”). Training increasingly takes place in simulation and with human motion data.
Key topics
- AI models
- Vision-language-action models translate instructions and camera images into movements. Below them, real-time controllers regulate every joint within milliseconds.
- Hardware
- Compact electric actuators, dexterous hands with a sense of touch and powerful batteries determine what a robot can physically achieve.
- Training
- Reinforcement learning in simulation, teleoperation and videos of human motion provide the data – transferring skills to the real world remains a core challenge.
- Scaling
- From prototype to series production: cost, reliability, maintenance and high-volume manufacturing are the next hurdles.
Key questions
- Which tasks can be solved robustly today – and which can’t?
- How long does a battery last in shift operation?
- How quickly can a robot learn a new task?
Leadership Leadership & strategy
Humanoid robots are not just an IT project. Deploying them changes processes, roles and work culture. Leadership determines whether demos turn into real value.
Key topics
- Choosing use cases
- Good starting points are clearly defined, recurring tasks in environments built for people – where specialised automation is too expensive or too rigid.
- Pilot → scale
- Start with measurable goals (throughput, quality, downtime, safety) and clarify early what must be in place to move into regular operation.
- Bringing people along
- Transparent communication, upskilling and early involvement of employees and works councils build acceptance.
- Economics
- Weigh the total cost (purchase or lease, integration, maintenance, software, training) honestly against benefits and risks.
Key questions
- Which problem are we solving – and is a humanoid the best solution?
- Who is responsible for the robot day to day?
- Which skills do our teams need?