Free public instrument from GAGE
The Physical AI Governance Atlas
As of September 2026 this ledger holds 44 records across 8 jurisdictions: 36 verified at a primary source, 1 reported by a secondary source, 0 announced with no document yet, 3 searched and absent, and 4 open questions. By kind: binding law 12, standard 13, guidance 5, program or strategy 10, open question 4.
- 44
- Records
- 36
- Verified at the source
- 3
- Announced or absent
- 4
- Open questions
8 jurisdictions, 64 primary sources
1 more reported, primary not reached
0 announced with no document, 3 searched and absent
Posed, sourced, not answered
As of 15 September 2026 the GAGE Physical AI Governance Atlas records 36 verified instruments, 1 reported at a secondary source, 0 announcements without a document, 3 absences and 4 open questions, across 8 jurisdictions. Counts are a floor, not a ceiling: an instrument the ledger has not found is not on it.
Why this ledger exists
The robot is governed in pieces, and the pieces are in different buildings
A humanoid on a factory floor is a machine to one regulator, an AI system to another, a workplace hazard to a third, a product with a liability regime to a fourth, and a recorder of hours to a fifth. No jurisdiction has one document for it, and the documents that exist were mostly written before a learned behavior could be the thing that fails.
This atlas keeps the record of every one of those documents, says which of the five objects each one actually governs, and marks where a jurisdiction has nothing. It sits beside the Robot Compliance Navigator, which maps a specific product's stack, and the Embodied AI Data Ledger, which follows the hours the robot learns from.
Every row is fetched at its source where the source could be reached, and says so where it could not. The verdict language is the discipline: a reader learns five words once and never has to guess what a row claims.
Verified
The document exists. The ledger fetched it at its publisher and quotes it.
Reported, primary not reached
A reliable secondary source carries it, and the primary document could not be reached. Printed with this label, never as verified.
Announced, no document yet
A body has said it will act. No document exists yet, so the row records the statement and nothing more.
Absent
The ledger searched and found no instrument. The record says where it looked and when.
Open question
No settled answer exists. The ledger poses the question, links the live debate, and does not answer it.
The grid
What the documents state, and where they are silent
For each of the five objects a robot document could govern, how many documents on the atlas state a rule and how many are silent. Silent on liability is the most common shape.
The machine
20 stated, 4 silent, 1 reported.
The AI inside it
15 stated, 11 silent, 1 open, 1 reported.
Liability
4 stated, 5 silent.
Workers and bystanders
15 stated, 0 silent, 1 open.
The data it records
4 stated, 4 silent, 1 reported.
Side by side
Every jurisdiction, counted by verdict and by kind
European Union
10 records
- Verified
- 7
- Reported
- 0
- Announced
- 0
- Absent
- 1
- Open question
- 2
6 binding law, 2 program or strategy, 2 open question.
United States
5 records
- Verified
- 4
- Reported
- 0
- Announced
- 0
- Absent
- 1
- Open question
- 0
1 binding law, 2 guidance, 2 program or strategy.
China
6 records
- Verified
- 5
- Reported
- 1
- Announced
- 0
- Absent
- 0
- Open question
- 0
3 standard, 1 guidance, 2 program or strategy.
Japan
3 records
- Verified
- 3
- Reported
- 0
- Announced
- 0
- Absent
- 0
- Open question
- 0
1 binding law, 2 program or strategy.
South Korea
3 records
- Verified
- 3
- Reported
- 0
- Announced
- 0
- Absent
- 0
- Open question
- 0
2 binding law, 1 program or strategy.
United Kingdom
3 records
- Verified
- 2
- Reported
- 0
- Announced
- 0
- Absent
- 1
- Open question
- 0
2 binding law, 1 guidance.
Germany
2 records
- Verified
- 2
- Reported
- 0
- Announced
- 0
- Absent
- 0
- Open question
- 0
1 guidance, 1 program or strategy.
Global
12 records
- Verified
- 10
- Reported
- 0
- Announced
- 0
- Absent
- 0
- Open question
- 2
10 standard, 2 open question.
Figures of record
Every number on this ledger, with who measured it and when
12 figures, each one printed in the unit its publisher used, beside the publisher and the date it was true. Nothing here is summed across sources, converted between units, or forecast.
- 318 plan items
Plan items returned by the platform for humanoid robot. China is drafting a five part humanoid robot safety standard, and the first robot control architecture standard took effect in 2026
National Public Service Platform for Standards Information, plan query for humanoid robot, primary source, as of .
- 0 standards
Published Chinese national standards with humanoid robot in the title. China is drafting a five part humanoid robot safety standard, and the first robot control architecture standard took effect in 2026
National Public Service Platform for Standards Information, plan query for humanoid robot, primary source, as of .
- 5 parts
Parts of the humanoid robot safety requirements national standard in drafting. China is drafting a five part humanoid robot safety standard, and the first robot control architecture standard took effect in 2026
National Public Service Platform for Standards Information, plan query for humanoid robot, primary source, as of .
- 28 percent
Growth in Chinese industrial robot output over the previous year, as the report prints it. China's 2026 work report names embodied intelligence a future industry and prints robot output growth
State Council Gazette 2026 number 9, Government Work Report delivered 5 March 2026, primary source, as of .
- 120 organisations, reported as over
Organisations Xinhua says took part in drafting the standard system. China announced a national standard system for humanoid robots and embodied AI, and the ministry's page could not be reached
People's Daily Online, carrying Xinhua, China releases national standard system for humanoid robotics and embodied AI, secondary source, as of .
- 76 members
Members of the first MIIT humanoid robot and embodied intelligence standards committee. China gave humanoid robots and embodied AI their own standards committee, with safety written into its remit
MIIT, announcement 2025 number 44 establishing MIIT/TC8, primary source, as of .
- 400 producers
European service robotics producers, as the Commission states it. The EU's robotics plan is one section of the Apply AI Strategy, and it funds uptake rather than rules
EUR-Lex, Communication COM(2025) 723 final, Apply AI Strategy, primary source, as of .
- 90,000 robots
Industrial robots installed in Europe in 2023, as the Commission states it. The EU's robotics plan is one section of the Apply AI Strategy, and it funds uptake rather than rules
EUR-Lex, Communication COM(2025) 723 final, Apply AI Strategy, primary source, as of .
- 50,000,000 yuan
Shanghai support ceiling per key technology project, as the plan prints it. Beijing and Shanghai each published an embodied AI industry plan, with targets and money attached
Shanghai Municipal Government, Implementation Plan for the Development of the Embodied AI Industry, primary source, as of .
- 50 enterprises, no fewer than
Beijing 2027 target for core enterprises in the chain, as the plan prints it. Beijing and Shanghai each published an embodied AI industry plan, with targets and money attached
Beijing Municipal Government, Embodied AI Science and Technology Innovation and Industry Cultivation Action Plan 2025 to 2027, primary source, as of .
- 223 pages
Pages in ISO 10218-2:2025. The world's industrial robot safety standard was reissued in 2025, in two parts, for the robot and for the cell
ISO, ISO 10218-2:2025 Robotics, Safety requirements, Part 2: Industrial robot applications and robot cells, primary source, as of .
- 404 responses
Written consultation responses received by the Law Commissions. The British statute came from a law reform project, and that report is the closest thing to a rationale a regulator will read
Law Commission, Automated Vehicles project page, primary source, as of .
The ledger
Every record, newest first
44 records. Each row opens a page carrying the answer, the verdict and what it means, the key facts, the figures with their sources, what it changes for a robot maker or a deployer, and the sources it was verified against.
Answers
What people ask the Physical AI Governance Atlas
Is there a law for robots?
Not one law anywhere on the atlas as of September 2026. 12 binding laws reach robots through machinery, product, AI, workplace or vehicle rules: South Korea's AI framework law never says robot, and its safety duty is triggered by compute (South Korea); South Korea has had a robot statute since 2008, and its robot certification articles were deleted in 2016 (South Korea); Japan wrote robot rules into occupational safety law in the 1980s, and they still govern every industrial robot there (Japan); Britain's one physical AI statute is about vehicles, and it makes an entity, not a driver, answer for the machine (United Kingdom); Britain has no Act with robot in its title, and the search that proves it is one line long (United Kingdom); and more. Each record says which of the five objects it governs.
Does the United States regulate humanoid robots?
No United States federal statute governs robots as such, and the agencies say so in their own words. Each absent record writes down where the atlas looked and when. What the United States does have is on the United States page: The nearest thing to a US federal robot programme is a measurement science project, and it makes test methods, not rules; America's AI Action Plan treats robotics as a manufacturing race and orders a supply chain convening, not a safety rule; The US regulates autonomous vehicles through exemptions and crash reporting, not through a self driving rule; US robot safety enforcement runs on general machine guarding plus a technical manual that points at private standards.
What has China published on humanoid robots?
6 records for China: China's 2026 work report names embodied intelligence a future industry and prints robot output growth (verified); Beijing and Shanghai each published an embodied AI industry plan, with targets and money attached (verified); China announced a national standard system for humanoid robots and embodied AI, and the ministry's page could not be reached (reported, primary not reached); China is drafting a five part humanoid robot safety standard, and the first robot control architecture standard took effect in 2026 (verified); China gave humanoid robots and embodied AI their own standards committee, with safety written into its remit (verified); China's humanoid robot policy is a 2023 MIIT document with dated milestones and a governance line inside it (verified). Records marked reported rest on a secondary source because the primary document could not be reached.
Who is liable when a learned behavior causes harm?
Nobody has settled it, which is why it is an open question on the atlas rather than a rule. 5 documents on the atlas govern robots while silent on liability; the records that do address it are on the liability dimension of the grid.
What binds a robot maker or deployer?
12 records of kind binding law are on the Physical AI Governance Atlas as of September 2026: South Korea's AI framework law never says robot, and its safety duty is triggered by compute; South Korea has had a robot statute since 2008, and its robot certification articles were deleted in 2016; Japan wrote robot rules into occupational safety law in the 1980s, and they still govern every industrial robot there; Britain's one physical AI statute is about vehicles, and it makes an entity, not a driver, answer for the machine; Britain has no Act with robot in its title, and the search that proves it is one line long; No United States federal statute governs robots as such, and the agencies say so in their own words; and 6 more. Each has its own page with the verdict, the facts, the sources it was checked against and the date.
What standard serves as evidence?
13 records of kind standard are on the Physical AI Governance Atlas as of September 2026: China announced a national standard system for humanoid robots and embodied AI, and the ministry's page could not be reached; China is drafting a five part humanoid robot safety standard, and the first robot control architecture standard took effect in 2026; China gave humanoid robots and embodied AI their own standards committee, with safety written into its remit; No published safety standard covers a legged humanoid, and the one that would is still a committee draft; The AI management system standard is the one certificate a robot maker can actually hold today; IEEE 7007 is an ontology for ethically driven robotics, and it is the only active standard of its kind; and 7 more. Each has its own page with the verdict, the facts, the sources it was checked against and the date.
What do regulators and agencies advise?
5 records of kind guidance are on the Physical AI Governance Atlas as of September 2026: China's humanoid robot policy is a 2023 MIIT document with dated milestones and a governance line inside it; Germany's practical robot rule is an accident insurers' handbook, and it is the document an inspector reaches for; The British statute came from a law reform project, and that report is the closest thing to a rationale a regulator will read; The US regulates autonomous vehicles through exemptions and crash reporting, not through a self driving rule; US robot safety enforcement runs on general machine guarding plus a technical manual that points at private standards. Each has its own page with the verdict, the facts, the sources it was checked against and the date.
What is the state funding or planning?
10 records of kind program or strategy are on the Physical AI Governance Atlas as of September 2026: Germany's federal robotics programme is a research ministry action plan with a booster for embodied AI, and it funds rather than regulates; Japan's robot adoption programme is a regional network run by the ministry, aimed at labour shortage rather than at rules; Japan is funding a domestic foundation model for physical AI, and calls the on site data its winning path; South Korea runs its humanoid push through a ministry led alliance, and the ministry names training data as a policy measure; China's 2026 work report names embodied intelligence a future industry and prints robot output growth; Beijing and Shanghai each published an embodied AI industry plan, with targets and money attached; and 4 more. Each has its own page with the verdict, the facts, the sources it was checked against and the date.
What has nobody settled?
4 records of kind open question are on the Physical AI Governance Atlas as of September 2026: Open question: how do you evidence a safety case for a controller that was learned rather than specified?; Open question: is a fleet wide behaviour update a new placing on the market?; Open question: is a humanoid in a home a machine, a consumer product or a care robot?; Open question: who is liable when a behaviour the robot learned causes the harm?. Each has its own page with the verdict, the facts, the sources it was checked against and the date.
What do the verdicts mean?
Verified: the document exists and the ledger fetched it at its publisher. Reported: a reliable secondary source carries it and the primary could not be reached. Announced: a body said it will act and no document exists. Absent: the ledger searched and found nothing, and the search is written into the record. Open: a question nobody has settled, posed and not answered.
How current is the Physical AI Governance Atlas?
Every record carries the date it was last verified; the ledger as a whole was last verified 16 September 2026 and holds 44 records with 64 primary sources. A change moves the record's own date and appears on the changelog, so a reader who cited a row can see whether it moved.
Every surface
Cut the ledger the way you need it
By jurisdiction
By kind
Every record page
- PAG-2026-0044: Germany's federal robotics programme is a research ministry action plan with a booster for embodied AI, and it funds rather than regulates
- PAG-2026-0043: Japan's robot adoption programme is a regional network run by the ministry, aimed at labour shortage rather than at rules
- PAG-2026-0042: Japan is funding a domestic foundation model for physical AI, and calls the on site data its winning path
- PAG-2026-0041: South Korea runs its humanoid push through a ministry led alliance, and the ministry names training data as a policy measure
- PAG-2026-0040: South Korea's AI framework law never says robot, and its safety duty is triggered by compute
- PAG-2026-0039: South Korea has had a robot statute since 2008, and its robot certification articles were deleted in 2016
- PAG-2026-0038: China's 2026 work report names embodied intelligence a future industry and prints robot output growth
- PAG-2026-0037: Beijing and Shanghai each published an embodied AI industry plan, with targets and money attached
- PAG-2026-0036: China announced a national standard system for humanoid robots and embodied AI, and the ministry's page could not be reached
- PAG-2026-0035: China is drafting a five part humanoid robot safety standard, and the first robot control architecture standard took effect in 2026
- PAG-2026-0034: China gave humanoid robots and embodied AI their own standards committee, with safety written into its remit
- PAG-2026-0033: China's humanoid robot policy is a 2023 MIIT document with dated milestones and a governance line inside it
- PAG-2026-0032: Open question: how do you evidence a safety case for a controller that was learned rather than specified?
- PAG-2026-0031: Open question: is a fleet wide behaviour update a new placing on the market?
- PAG-2026-0030: Open question: is a humanoid in a home a machine, a consumer product or a care robot?
- PAG-2026-0029: Open question: who is liable when a behaviour the robot learned causes the harm?
- PAG-2026-0028: No published safety standard covers a legged humanoid, and the one that would is still a committee draft
- PAG-2026-0027: The AI management system standard is the one certificate a robot maker can actually hold today
- PAG-2026-0026: IEEE 7007 is an ontology for ethically driven robotics, and it is the only active standard of its kind
- PAG-2026-0025: UL 4600 is the one published standard that asks an autonomous product for a safety case
- PAG-2026-0024: The standard for harm with no broken part is SOTIF, written for road vehicles and borrowed by everyone else
- PAG-2026-0023: Two functional safety standards decide whether a robot's stop is trustworthy, and both predate learned control
- PAG-2026-0022: Autonomous mobile robots in a warehouse are governed as driverless industrial trucks, by name
- PAG-2026-0021: The only safety standard written for robots that live with people is from 2014 and is marked for revision
- PAG-2026-0020: The collaborative robot document is a 2016 technical specification, and its replacement is still at proposal stage
- PAG-2026-0019: The world's industrial robot safety standard was reissued in 2025, in two parts, for the robot and for the cell
- PAG-2026-0018: Japan wrote robot rules into occupational safety law in the 1980s, and they still govern every industrial robot there
- PAG-2026-0017: Germany's practical robot rule is an accident insurers' handbook, and it is the document an inspector reaches for
- PAG-2026-0016: The British statute came from a law reform project, and that report is the closest thing to a rationale a regulator will read
- PAG-2026-0015: Britain's one physical AI statute is about vehicles, and it makes an entity, not a driver, answer for the machine
- PAG-2026-0014: Britain has no Act with robot in its title, and the search that proves it is one line long
- PAG-2026-0013: No United States federal statute governs robots as such, and the agencies say so in their own words
- PAG-2026-0012: The nearest thing to a US federal robot programme is a measurement science project, and it makes test methods, not rules
- PAG-2026-0011: America's AI Action Plan treats robotics as a manufacturing race and orders a supply chain convening, not a safety rule
- PAG-2026-0010: The US regulates autonomous vehicles through exemptions and crash reporting, not through a self driving rule
- PAG-2026-0009: US robot safety enforcement runs on general machine guarding plus a technical manual that points at private standards
- PAG-2026-0008: The EU has no dedicated robotics strategy document, only a research policy page and a section of the AI strategy
- PAG-2026-0007: The EU's robotics plan is one section of the Apply AI Strategy, and it funds uptake rather than rules
- PAG-2026-0006: EU product safety law already names learning and predictive functions as a safety factor for consumer robots
- PAG-2026-0005: The EU withdrew its AI Liability Directive, so there is no AI specific fault regime for a robot that harms someone
- PAG-2026-0004: The revised EU product liability rules make software a product and reach a robot that changes after it is sold
- PAG-2026-0003: A robot that talks to people owes an AI Act disclosure, and a robot that reads emotions owes another
- PAG-2026-0002: The Digital Omnibus took AI enabled machinery out of the AI Act high risk lane and handed it to the Machinery Regulation
- PAG-2026-0001: The EU Machinery Regulation applies from 14 January 2027 and puts self evolving safety functions in the third party lane
Take the data
The whole dataset, free, in two formats
Licensed CC BY 4.0. Use it in an article, a paper, a slide or a product. The only condition is attribution, and the citation page gives you the line to paste.
- ledger.jsonEvery field of every record, the shape documented on the data page.
- ledger.csvOne row per record, figures and facets flattened, for a spreadsheet or a stats package.
How the ledger is built, what the verdicts mean, and what the gate refuses: the method page. Every change, dated: the changelog. The kinds on the shelf: binding law, standard, guidance, program or strategy, open question. Something missing or wrong is a bug, and we want to hear about it. Which of these binds your product is answered by the Robot Compliance Navigator. The hours the robot learns from have their own record on the Embodied AI Data Ledger. The jobs that name these standards are counted on the Physical AI Hiring Index.
Cite this page
Free to reuse under CC BY 4.0, with attribution.
- In a sentence
- According to the GAGE Physical AI Governance Atlas (as of 15 September 2026), physical ai governance atlas.
- APA
- GAGE (Global Academy of Generative-AI Education). (2026). Physical AI Governance Atlas. Physical AI Governance Atlas. Retrieved 15 September 2026, from https://www.gage.academy/tools/physical-ai-governance-atlas
- MLA
- "Physical AI Governance Atlas." Physical AI Governance Atlas, GAGE (Global Academy of Generative-AI Education), 15 September 2026, https://www.gage.academy/tools/physical-ai-governance-atlas.
- Chicago
- GAGE (Global Academy of Generative-AI Education). "Physical AI Governance Atlas." Physical AI Governance Atlas. Last modified 15 September 2026. https://www.gage.academy/tools/physical-ai-governance-atlas.
- Permalink
- https://www.gage.academy/tools/physical-ai-governance-atlas
Last updated . Every record re verified . The ledger is checked monthly, first Monday, and the same day for any new law or standard publication.
318 plan items
GAGE briefings tell you which AI regulation deadlines are coming, what they actually require of you, and when a program opens.