Beyond Standards: Why Humanoid Robot Safety Requires What No Framework Has Codified Yet
This article was written by Gokul Krithivasan, Managing Partner and Co-Founder of SRES. Gokul is a senior safety expert with over 14 years of experience supporting OEMs, suppliers, and technology companies in the safety assurance of complex autonomous and AI-enabled systems across automotive, autonomy, logistics, and robotics applications.
Looking to go deeper? SRES provides expert-led training for Physical AI systems, along with hands-on consulting support, including robotics safety, to help organizations develop and deploy Physical AI systems responsibly.
The humanoid robot revolution is happening right now. Companies are racing to deploy bipedal, wheeled, AI-driven workers into warehouses, manufacturing hubs, and even our homes.
Recent public incidents highlight why safety must evolve alongside deployment. In June 2026, a humanoid robot performing a martial arts demonstration accidentally struck a child during a public event. While the child was not seriously injured and the root cause was not publicly disclosed, the incident has sparked widespread discussion about how humanoid robots should be evaluated when operating in environments shared with people.
As humanoid systems move from research labs into factories, warehouses, construction sites, and public spaces, the industry faces a new class of safety challenges that extend beyond traditional hardware and software failures.
At SRES, we are deeply embedded in the humanoid robotics ecosystem — actively collaborating with humanoid robot manufacturers and component suppliers, while training independent certification bodies on evaluating the overall safety of safety-critical Physical AI systems.
Through these deep technical engagements, we keep having déjà vu. Traditional robotics functional safety frameworks and standards are necessary but not sufficient for developing and scaling a truly safe humanoid robot. They are unequipped to address the behavioral safety challenges emerging with complex AI-based systems. A simple glare from a warehouse window or an unfamiliar object geometry in the kitchen can create a perception insufficiency. The code runs perfectly and the hardware has no faults, yet the resulting physical behavior is unsafe.
Right now, humanoid manufacturers are asking the exact same foundational safety questions that self-driving car companies grappled with in the early 2010s while implementing ISO 26262. The good news? These are not uncharted mysteries. The frameworks already exist, refined over a decade of high-stakes deployments on public roads, and they can be adapted for humanoids in both factories and homes.
The Limits of the Legacy Functional Safety Playbook
For decades, industrial robotics safety has been a well-understood, heavily codified discipline. Traditional robot safety consultants are very proficient in frameworks like IEC 61508, the Machinery Directive, and various UL standards. Their expertise focuses primarily on Functional Safety — ensuring that if a hardware part fails or a line of code crashes, the system defaults to a predictable, safe state.
But humanoid robots do not live behind physical safety cages or light curtains. They operate in dynamic, semi-structured spaces, relying on deep learning AI models for perceiving and interacting with their environment.
Because of this, legacy safety frameworks are behind the curve. AI-based systems can misbehave even in the absence of a hardware or software fault. Every component can operate exactly as specified, yet the systemic outcome can still be unsafe.
Say a household humanoid butler is carrying hot coffee when a family pet darts across the room. What if the robot’s vision system confuses the pet’s fast-moving shadow with a physical barrier, triggering an immediate, hard stop? The abrupt deceleration causes the coffee to spill, leading to spilled coffee at best, a severe burn at worst — triggered by a routine, benign event.
Traditional functional safety cannot fix this — we need a shift toward behavioral safety.
Borrowing from the ADS Playbook: Enter SOTIF
When Advanced Driver Assistance Systems (ADAS) and Automated Driving Systems (ADS) began scaling up in the 2010s, the automotive industry quickly realized that safety of complex autonomous systems required a more holistic approach for product safety — including functional safety, behavioral safety, and operational safety. We explored this in our 2023 piece “ISO 26262 is like bringing a knife to a gun fight with AI”.
To solve this, the industry developed ISO 21448, known as SOTIF (Safety of the Intended Functionality). Rather than focusing only on failures, SOTIF addresses functional insufficiencies and environmental triggering conditions through rigorous, scenario-based validation.
Rather than reinventing the safety wheel, the humanoid robotics sector can and should borrow and adapt this ADAS/ADS playbook. We are already starting to see these concepts reflected in the upcoming ISO/IEC TS 22440 standard, where system misbehaviors must be identified in the hazard and risk analysis (HARA), alongside traditional functional safety considerations.
Time-to-Market: Why Humanoid Companies Can’t Afford a Ramp-Up
As the humanoid market accelerates, companies do not have the luxury of waiting for traditional robotics safety approaches to slowly adapt to behavioral safety and non-deterministic AI architectures. Time-to-market and commercial viability depend on getting safety right now.
At SRES, we aren’t adapting to this paradigm; we helped pioneer it. Over the last 15 years, our consultants have designed, analyzed, and assessed behavioral safety concepts across autonomous systems — including robotaxis, autonomous trucking, delivery robots, and industrial autonomous ground vehicles (AGV).
The race to deploy humanoid robots won’t be won by checking boxes. It will be won by building sophisticated behavioral safety concepts on top of a strong functional safety and quality engineering foundation.
Have insights or questions? Send us an email at info@sres.ai or leave a comment below—we welcome thoughtful discussion from our technical community.
Need support for humanoid robotics or advanced Physical AI systems? SRES supports organizations with integrated functional safety, autonomy safety, cybersecurity, and AI assurance consulting. [Learn more]
