AI Policy & Governance: The Rules of the Road

Humanoid robots are no longer just science fiction. They're working in factories, delivering packages, and soon they might be helping in our homes. As these human-like machines become more common, governments and safety organizations worldwide are scrambling to create rules that keep everyone safe while allowing innovation to flourish.

Think of it like the early days of automobiles—when cars first appeared on roads designed for horses, new traffic laws, safety standards, and licensing requirements had to be created. Today, we're facing a similar challenge with humanoid robots.

The Regulatory Landscape: Multiple Layers of Protection

The rules governing humanoid robots come from several sources, creating layers of protection that help ensure these machines operate safely in our world.

International and Regional Laws

The European Union leads the charge with the AI Act, which became law in August 2024. This groundbreaking legislation treats AI systems like humanoid robots based on their risk level. High-risk robots used in factories or public spaces must meet strict safety requirements, while low-risk applications like entertainment robots face fewer restrictions. The Act categorizes AI systems into four risk levels: unacceptable, high, limited, and minimal, with each level carrying distinct requirements to ensure safety, transparency, and accountability.

The EU AI Act prohibits certain AI applications that threaten citizens' rights, including biometric categorization systems based on sensitive characteristics and emotion recognition in workplaces and schools. Robots that could pose physical risks to humans fall under transparency measures, and human oversight requirements.

In China, Shanghai has published the world's first humanoid robot governance guidelines, emphasizing that these machines must "safeguard human dignity" and "not threaten human security". These guidelines require companies to establish risk warning procedures and emergency response systems while providing training on ethical and lawful use of humanoid robots.

In the United States, the approach is more collaborative. The National Institute of Standards and Technology (NIST) provides a framework through its AI Risk Management Framework that companies can follow voluntarily, focusing on four key functions: Govern (set up oversight), Map (identify risks), Measure (test safety), and Manage (control ongoing operations). This framework helps organizations responsibly design, develop, implement, and use AI systems with a focus on ethical and risk-aware implementation.

Safety Standards: The Technical Rulebook

The most important safety standard is ISO 10218, which was completely updated in 2025 after nearly eight years of work. This standard serves as the technical manual that tells engineers exactly how to build safe industrial robots. The 2025 version includes new requirements for cybersecurity and reflects how robots now work alongside humans in collaborative applications.

The updated ISO 10218 features extensive revisions that focus on making functional safety requirements more explicit rather than implied. This shift enhances clarity and usability, making compliance more straightforward for manufacturers and integrators alike. Key updates include clarified functional safety requirements, integrated safety requirements for collaborative robot applications that consolidate the previously separate ISO/TS 15066, new robot classifications with corresponding functional safety requirements, and cybersecurity requirements pertaining to industrial robot safety.

The standard now encompasses approximately 38 safety functions, with requirements that robots meet specific performance levels (PLd Category 3 or SIL2 HFT=1) for critical safety systems. These functions cover everything from how fast a robot can move near humans to what happens if something goes wrong, providing detailed safety instructions for every possible scenario.

Workplace Safety Rules

The Occupational Safety and Health Administration (OSHA) in the U.S. has updated its robot safety guidelines to align with international standards. In 2022, OSHA, in collaboration with the National Institute for Occupational Safety and Health (NIOSH) and the Association for Advancing Automation, updated the OSHA Technical Manual chapter on Industrial Robot Systems and Industrial Robot System Safety.

These updated guidelines include technical information on hazards associated with industrial and emergent robot applications, safety considerations for employers and workers, and risk assessments and risk reduction measures. The manual serves to guide OSHA compliance officers during inspections at facilities with robotic systems and provides a technical resource for safety professionals overseeing robotic systems in workplaces.

How It Works in Practice: The Compliance Journey

When a company wants to deploy humanoid robots, they must follow a structured process that ensures safety and regulatory compliance at every step.

1. Risk Assessment
   Companies identify what their robot will do and what could go wrong. For example, if a humanoid robot will work in a warehouse moving boxes near human workers, engineers must consider risks like collisions, equipment malfunctions, or cybersecurity breaches. The NIST AI RMF provides a systematic approach where organizations must "Map" their AI risks, identifying potential hazards specific to their intended use cases.
2. Safety Testing
   The updated ISO standards require specific tests for force limits, speed controls, and safety systems. Robots must prove they can stop immediately if they detect a human nearby or if something unexpected happens. The new standards include test methodologies for different robot classifications, with Class I robots having low energy transfer capability requiring specific testing protocols.
3. Documentation and Certification
   Companies must create detailed records showing how their robot meets safety standards. In Europe, this includes producing a "Declaration of Conformity"—essentially a legal document stating the robot is safe for its intended use. The documentation must include information about functional safety characteristics, risk management procedures, and ongoing monitoring plans.
4. Ongoing Monitoring
   The rules don't stop once a robot starts working. Companies must continuously monitor performance, log incidents, and update safety measures as needed. The NIST framework emphasizes the "Manage" function, requiring organizations to regularly review and adjust their risk management strategies based on real-world performance data.

Real-World Examples: Robots Following the Rules

Tesla's Optimus: Privacy-First Design

Tesla's humanoid robot Optimus demonstrates how privacy regulations are being built into robot design from the start. The company processes data locally on the robot rather than sending personal information to remote servers, addressing privacy concerns before they become problems. Recent developments show Optimus learning new tasks by watching YouTube videos, with Tesla emphasizing that training videos undergo "rigorous filtering to block out any violent, illegal or inappropriate content".

However, questions about Tesla's transparency have emerged, as seen in Robotics Expert: Tesla's Demos Hide This, where experts analyze the technical reality behind Tesla's robot demonstrations and discuss the importance of transparency in robotics development.

Figure 01: Transparent Decision-Making

Figure AI's robot showcases how modern humanoids document their decision-making process. When the robot decides to pick up an object or move in a certain direction, it creates logs that regulators and safety investigators can review later. Figure Status Update - OpenAI Speech-to-Speech Reasoning demonstrates the robot's ability to understand context, make decisions, and explain its reasoning process in real-time interactions with humans.

Mercedes-Benz and Apollo: Factory Integration

Mercedes-Benz is testing Apptronik's Apollo robot in their facilities, following strict ISO 10218 protocols. The robots work within clearly defined safety zones, with speed and distance controls that protect human workers. Mercedes-Benz has invested a "low double-digit million-euro sum" in Apptronik and is conducting trials at its Digital Factory Campus in Berlin and in Hungary.

The practical implementation of Apollo in Mercedes-Benz factories is shown in New Mercedes is testing the use of Apollo, a humanoid robot from Apptronik, in production, demonstrating how humanoid robots are being integrated into automotive manufacturing while following safety protocols and regulations.

Unitree G1: Affordable Safety

Even lower-cost robots like Unitree's $16,000 G1 robot include safety features like LiDAR sensors for detecting humans and detailed logging systems to track performance and incidents. The G1 weighs just over 100 pounds, stands 4 feet 2 inches tall, and can walk at speeds of roughly 4.5 mph with a battery life of about 2 hours.

The impressive capabilities of the Unitree G1 are showcased in Brand new video from Unitree Robotics – 'Kungfu BOT: Unitree G1', demonstrating the robot's advanced mobility, balance, and ability to perform complex movements while maintaining safety protocols.

Beyond Safety: Ethics and Rights

Regulators aren't just concerned about physical safety. They're also grappling with broader questions that touch on fundamental aspects of human society and robot integration.

Privacy and Data Protection

As robots become equipped with cameras, microphones, and sensors, protecting personal information becomes critical. The EU AI Act includes specific provisions for data governance and privacy-enhanced technologies, requiring that AI systems leverage data minimization to protect sensitive user data. Companies must implement privacy-enhancing technologies and ensure transparency about what data is collected and how it's used.

Job Displacement and Economic Impact

The integration of humanoid robots raises concerns about workforce displacement. Some European lawmakers have discussed robot taxation similar to proposals by French presidential candidate Benoît Hamon, who suggested companies replacing workers with machines should face higher charges to offset the incentives to remove humans from the workforce. This reflects broader discussions about how to manage the economic transitions that come with increased automation.

Bias and Fairness

Ensuring robots treat all people fairly is a growing concern. The NIST AI RMF emphasizes the importance of managing harmful bias, identifying three categories that must be addressed: statistical bias, cognitive bias, and systemic bias. The framework requires organizations to implement measures that promote equality and equity through the removal of discrimination.

Accountability and Responsibility

Determining who's responsible when a robot makes a mistake remains a complex challenge. Current frameworks generally place responsibility on the human operators, manufacturers, or system integrators rather than the robots themselves. The EU AI Act establishes clear liability chains, with providers and deployers of AI systems bearing different levels of responsibility depending on their role in the system's lifecycle.

The ongoing debate about robot rights and accountability continues in academic and policy circles, as explored in Will Robots Ever Have Rights?, where political science professors examine arguments about robot rights and the human tendency to anthropomorphize non-human entities.

Some European lawmakers have even discussed whether advanced robots should have legal rights similar to corporations, though this remains controversial and theoretical. The European Parliament previously drafted a motion that would define robots as "electronic people," though this concept has not been adopted into law.

The Path Forward: Innovation Within Guardrails

The goal of these regulations isn't to stop progress—it's to ensure that humanoid robots develop safely and responsibly. Companies that follow these emerging rules early often find it easier to expand globally and gain public trust.

Evolving Regulatory Frameworks

As robots become more sophisticated and widespread, we can expect to see more specific rules for different industries. Healthcare applications will likely require different standards than transportation or home care robots. The modular approach being taken by organizations like ISO allows for industry-specific adaptations while maintaining core safety principles.

International Coordination

There's growing recognition of the need for international coordination to avoid regulatory conflicts. The Shanghai guidelines specifically call for "establishing a global governance framework" and an international think tank dedicated to governing humanoid robots. This global approach could help ensure that safety standards are consistent across borders while allowing for regional variations in implementation.

Regular Updates and Evolution

Technology standards require regular updates as technology advances. The nearly eight-year gap between ISO 10218 updates demonstrates the challenge of keeping pace with rapid technological development. Future frameworks will likely need more agile update mechanisms to address emerging technologies and use cases.

Greater Public Input

There's increasing recognition that robot policies should include broader public input, not just technical experts and industry stakeholders. The NIST AI RMF was developed through collaboration with over 240 entities from public and private sectors as well as academia, demonstrating the value of diverse perspectives in developing comprehensive frameworks.

What This Means for You

Whether you're a business owner considering robotic workers, an employee who might work alongside robots, or simply someone curious about the future, these regulations affect you directly. They're designed to ensure that as humanoid robots become part of our daily lives, they do so safely and with respect for human rights and dignity.

For Businesses

Companies planning to deploy humanoid robots should start familiarizing themselves with relevant standards now. Early compliance not only ensures safety but can also provide competitive advantages as regulations become more stringent. The structured approach provided by frameworks like ISO 10218 and NIST AI RMF offers clear pathways for responsible development and deployment.

For Workers

Understanding these regulations can help workers know what protections they have when working alongside robots. The emphasis on human-robot collaboration rather than replacement suggests that many jobs will evolve rather than disappear, but workers should prepare for changing skill requirements and new forms of human-machine interaction.

For Society

These regulations represent society's attempt to shape how technology integrates into our lives. Public awareness and engagement with these policy discussions can help ensure that regulatory frameworks reflect broader societal values rather than just technical or commercial considerations.

The regulatory landscape for humanoid robots is complex and rapidly evolving, but the core principle remains simple: robots should help humans thrive, not put them at risk. By establishing clear rules now, we're laying the groundwork for a future where humans and humanoid robots can work together safely and effectively.

Looking toward the future of human-robot collaboration, The Future of Human Robot Collaboration explores the exciting possibilities and potential benefits of humans and robots working together, highlighting how regulatory frameworks are enabling safe and productive partnerships between people and machines.

As we continue to navigate this technological transformation, the success of these regulatory efforts will ultimately be measured not just by their ability to prevent harm, but by how well they enable the beneficial applications of humanoid robotics while preserving human agency, dignity, and wellbeing. The rules of the road for humanoid robots are still being written, but they're guided by the fundamental principle that technology should serve humanity's best interests.

Media Gallery