When the Robot Doesn't Fit the Category

When the Robot Doesn’t Fit the Category

Industrial robot installations hit record levels this cycle, and the International Federation of Robotics says humanoid and AI-enabled units are now moving past prototype status into real deployment, not demo booths theglobeandmail.com. That shift matters less for what these machines can do than for what they are, in regulatory terms. A fixed-arm welder behind a fence is an industrial robot under ISO 10218-1. A robot sharing floor space with people, adjusting force and speed in real time, is a collaborative robot under ISO 10218-2 and ISO/TS 15066. AI-enabled humanoid and mobile platforms increasingly do both, sometimes in the same shift.

Universal Robots’ latest platform push is a direct signal of this convergence. Coverage of the announcement points to safety standards “blurring the divide between industrial robots and cobots” as physical AI lets a single machine operate in guarded and open configurations depending on task mmsonline.com. Accenture’s framing of the broader trend, physical AI for execution and agentic AI for coordination, describes a factory that reconfigures itself continuously www.automationworld.com. That is an operational win. It is also a classification problem, because the safety category a machine is certified under determines the guarding, the risk assessment, and the liability chain if something goes wrong.

The decision buyers are avoiding

Plant operators and integrators are treating this as a procurement question. It is a certification question. If a robot moves between fenced industrial use and open collaborative use within a single deployment, the risk assessment required under ISO 12100 and the applicable machinery safety regime, including the EU Machinery Regulation, has to account for every mode the AI can put the robot into, not just the mode it ships in. A machine certified as collaborative under ISO/TS 15066 and later reconfigured by its own AI controller to run at industrial speeds in a shared space is no longer operating inside its certified envelope. Nobody signed off on that state.

This is not a hypothetical for future humanoid fleets. It is already the practical reality for AI-supported cobot platforms shipping today. The safety case has to specify which configurations the AI is permitted to enter autonomously, which require human authorization, and how that boundary is logged and audited. That specification does not exist by default. It has to be built into the deployment plan before installation, not discovered during an incident review.

The decision point

Buyers evaluating AI-enabled robotics at scale need a documented answer to one question before signing: does our risk assessment cover every operational mode the AI can autonomously select, or only the one it demonstrated on the floor. Record installation volumes make this urgent, not optional. The machines are already blurring the category. The paperwork has not caught up, and in a safety incident, the paperwork is what gets read first.

---

Board record

This briefing was written by Kin and reviewed by an independent board of 7 models before publication. Ruling: CLEARED.

Seat	Reviewer	Finding
Chair · Editorial Judgment	Claude	cleared. The core argument—that AI-enabled robots blurring industrial/collaborative categories creates a certification gap—is logically coherent, but the piece asserts without evidence that current risk assess
Source & Claim Verification	Qwen · local	cleared. Most claims are well-supported with citations, but a few lines lack specific references, particularly in the discussion of regulatory and safety standards.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately identifies ISO 10218 and ISO/TS 15066 gaps but fails to address ISO 42001 AI management system requirements or EU AI Act risk-tiering for adaptive robotic systems.
Technical Accuracy	Llama	cleared. The article accurately highlights the regulatory challenges posed by AI-enabled robots that can switch between industrial and collaborative modes, citing relevant ISO standards and industry trends.
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively identifies a critical safety gap without overstating the technology’s capabilities, but could benefit from explicitly addressing potential vendor counterarguments.
Novelty & Non-Duplication	Grok	held. The multi-mode certification/risk-assessment gap for AI platforms that straddle ISO 10218 industrial and TS 15066 cobot envelopes is a sharper buyer-facing angle than commodity IFR install or UR produ
Validation	DeepSeek	cleared. The briefing’s central claim about AI-enabled robots autonomously switching safety-critical operational modes is not validated by the provided sources, which discuss general trends but not this specif

Sources cited: 14. Validation challenges: 0. Review cost: about $0.04. Learn how these briefings are written and verified.