Robot Governance in Practice: Safety Assessment, Public Oversight, and Japan's Delivery Robot Framework

1. Introduction: From Technological Capability to Institutional Responsibility

In 2026, three influential technology figures offered different perspectives on artificial intelligence's future. Elon Musk questioned whether humans could retain control over intelligence vastly superior to their own. Bill Gates warned about AI misuse, cybersecurity threats, and social disruption. Sam Altman observed that technological capabilities had advanced faster than their economic and organizational effects. ^[1][2][3]

These concerns involve different risks and timescales. The strongest version of Musk's hypothetical loss-of-control scenario has no established institutional solution. Gates's warnings concern harmful uses of powerful systems, while Altman's observations highlight the difficulties of technological adoption. At the Commonwealth Bank of Australia's Accelerate AI event, CEO Matt Comyn also acknowledged growing uncertainty and anxiety among employees as AI adoption progressed. ^[4]

For physical robots, these broader concerns lead to a more immediate question: how should institutions organize responsibility for autonomous machines operating in environments shared with people?

Robotic systems are already subject to machinery safety standards, product regulations, traffic rules, and organizational responsibilities. Their deployment involves manufacturers, operators, technical assessment organizations, regulators, and individuals affected by their operation.

The challenge is therefore not simply to establish new governance arrangements, but to understand how existing ones work together.

Japan's low-speed delivery robot framework offers a useful case. Since April 2023, qualifying remotely operated small vehicles have been permitted to travel on public roads through a notification procedure. Technical assessment, road-use administration, and operational supervision are distributed among different actors.

This article examines that division of responsibilities, focusing particularly on changes after initial assessment. Software updates, new operating environments, and increases in the number of robots supervised by one operator may alter the conditions under which safety was originally evaluated.

The central argument is that effective robot governance depends not only on assigning responsibilities before deployment, but also on maintaining reliable connections between technical assessment, actual operation, significant changes, and regulatory oversight.

2. Existing Safety Standards and Their Boundaries

Robot governance operates within an established landscape of technical safety standards.

ISO 12100:2010 provides general principles for machinery risk assessment and risk reduction, including methods for identifying hazards and evaluating risks during relevant phases of a machine's lifecycle. ^[5]

More specialized standards address particular robot categories and operating environments.

ISO 13482:2014 establishes safety requirements for defined categories of personal care robots, including mobile servant robots, physical assistant robots, and person carrier robots. Its applicability depends on the robot's intended function and the scope of the standard rather than the mere presence of autonomous navigation. ^[6]

ISO 3691-4:2023 addresses driverless industrial trucks and their systems, including certain autonomous mobile robots used for industrial load handling. However, the standard excludes additional hazards associated with trucks intended to operate in public zones and with operation on public roads. ^[7]

Consequently, compliance with ISO 3691-4 does not, by itself, establish that an industrial mobile robot is suitable for delivery operations on public sidewalks.

ISO 31101:2023 approaches safety from a different perspective. It establishes requirements for safety management systems covering application services provided by service robots in unstructured human environments, including certain delivery applications. It is not intended to serve as a product safety standard. ^[8]

These standards address related but distinct responsibilities: machinery design, product-specific hazards, and the management of robotic services. None should automatically be treated as a complete legal or operational framework for every public-road delivery robot.

2.1 International standardization for autonomous delivery robots

International standardization is also developing requirements more directly related to autonomous cargo transport.

Within the International Electrotechnical Commission (IEC), Technical Committee 125 established Working Group 6 to develop IEC 63281-2-2, E-Transporters — Part 2-2: Safety requirements and test methods for autonomous cargo e-Transporters.

According to materials presented to Japan's Ministry of Economy, Trade and Industry (METI) in March 2025, development had continued since November 2022, and the project entered the Committee Draft for Vote (CDV) stage in January 2025. ^[9]

The Robot Delivery Association (RDA) serves as the Japanese domestic deliberation body for IEC/TC125.

European standardization records have also listed the corresponding prEN IEC 63281-2-2:2025 document as a draft. Final publication of IEC 63281-2-2 as an International Standard could not be independently confirmed for this review.

The project nevertheless demonstrates that international technical institutions are addressing safety requirements specifically relevant to autonomous cargo transporters.

For robot governance, the important distinction is between the existence of a standardization initiative and the applicability of a published standard.

Technical standards evolve alongside deployment practices. Their relationship with national operating rules and conformity-assessment arrangements must therefore be examined according to the system and activity involved.

3. Japan's Delivery Robot Framework: A Division of Responsibilities

Japan provides a concrete example of robot governance organized through public law, private technical assessment, and operational responsibility.

Amendments to the Road Traffic Act effective from 1 April 2023 established a framework for remotely operated small vehicles (遠隔操作型小型車).

Qualifying vehicles may travel on public roads under specified conditions after advance notification to the relevant prefectural public safety commission.

This is a notification system (届出制), rather than a general discretionary operating-permit system. ^[10]

The framework distinguishes the technical characteristics of the vehicle from the administrative supervision of its road use.

3.1 What the Road Traffic Act regulates

Qualifying vehicles are subject to dimensional and operational requirements.

The prescribed maximum dimensions are generally 120 cm in length, 70 cm in width, and 120 cm in height, subject to specified exclusions for certain components. Vehicles must use an electric motor, be structurally incapable of exceeding 6 km/h, and possess emergency stopping equipment meeting the applicable requirements. ^[11]

The vehicle must also support actual remote operation. The police administrative interpretation explains that the ability to activate an emergency stop alone does not satisfy the requirement for remote control of vehicle movement. ^[12]

Before remotely operating a qualifying vehicle on public roads, the user must generally submit the required notification at least one week in advance.

The notification identifies matters including the operator, intended operating locations, remote-operation facilities, staffing arrangements, transportation activities, emergency stopping equipment, and prescribed vehicle specifications.

Changes to legally specified notification matters also require notification.

The Road Traffic Act further provides public safety commissions and police authorities with powers concerning reports, inspections, traffic management, and necessary intervention within the applicable statutory conditions.

These provisions establish an important part of the governance structure. Authorities can identify who is operating a vehicle, where it is intended to travel, and how its remote operation is organized.

However, the notification procedure does not make the police the technical safety certification body for the robot.

3.2 Private safety assessment and public-road notification

A distinctive feature of Japan's framework is its treatment of technical safety evidence.

Under Article 5-4(3)(iv) of the Road Traffic Act Enforcement Regulations, the notification must be accompanied by documentation describing the vehicle's structure and performance.

One accepted form is a certificate issued following an assessment conducted by a suitably qualified general incorporated association or foundation.

Other objective technical documentation, including manufacturer-prepared specifications, may also be accepted. ^[11][12]

The police administrative circular explains that the purpose of this requirement is to enable public safety authorities to understand the characteristics of vehicles operating on roads so that they can exercise their traffic-management and enforcement responsibilities. It is not intended to require the public safety commission to conduct its own technical safety examination.

The Robot Delivery Association operates a voluntary conformity-assessment scheme for remotely operated small vehicles. The assessment covers the robot, its remote-operation equipment, and the integrated system. ^[13]

A certificate issued through this process can serve as technical evidence accompanying a road-use notification.

However, third-party conformity assessment is not a universal mandatory prerequisite for submitting the notification. The legal framework also accepts other qualifying objective documentation.

This distinction matters because the two documentation routes provide different forms of assurance.

A third-party conformity assessment introduces technical evaluation by an organization separate from the manufacturer. Manufacturer-prepared specifications can provide relevant information about a robot's structure and performance, but their submission does not necessarily involve independent evaluation.

Neither route changes the underlying division of responsibilities.

Manufacturers and operators remain responsible for their relevant technical and operational obligations. Private assessment organizations evaluate systems within the scope of their schemes. Public authorities supervise road use and exercise the powers established by law.

3.3 What an RDA conformity certificate actually covers

The scope of the RDA assessment is important for understanding the relationship between initial evaluation and subsequent operation.

According to the association's published assessment overview, the evaluation concerns specific robot and remote-operation configurations under defined conditions. Unless otherwise stated, the assessment assumes conditions such as clear or cloudy weather, dry road surfaces, temperatures between 5°C and 35°C, and the absence of strong winds or visibility restrictions. ^[13]

The document also identifies matters outside the assessment's scope.

These include comprehensive safety under every possible operating condition, the manufacturer's production and quality-management systems, malicious interference by operators or third parties, and certain product-safety characteristics not covered by the relevant criteria.

Consequently, a conformity certificate should be understood as evidence of assessment against specified criteria and assumptions, rather than a guarantee of comprehensive safety in every environment.

The same document addresses changes after assessment.

Changing the combination of an assessed robot and its remote-operation equipment requires reassessment. Where the submitted operating environment or operating arrangements change, additional assessment of the relevant differences is required under the scheme.

This introduces an important lifecycle consideration.

The assessment is connected to a particular system configuration and set of operating assumptions. Continued reliance on its results depends on whether those assumptions remain applicable.

A certificate may therefore provide valuable technical assurance at deployment while leaving manufacturers and operators with continuing responsibilities to identify changes affecting safe operation.

3.4 How the framework operates in practice

Aichi Prefecture's delivery robot demonstration provides an example of these arrangements working together.

In December 2025, the prefecture announced a project using ROBO-HI's DeliRo delivery robot in Nagoya's Sakae district.

The project used the public-road notification framework and involved autonomous navigation supported by remote monitoring and operation.

The prefecture also stated that the robot had passed the Robot Delivery Association's safety-standard assessment. ^[14]

The project illustrates three institutional functions.

The private assessment addressed the robot's conformity with the applicable safety criteria.

The notification procedure established information about the intended road operation and responsible operator.

Public authorities retained their statutory powers concerning road traffic and necessary intervention.

These functions are complementary rather than interchangeable. An assessment certificate does not constitute unrestricted permission to operate anywhere, while completion of the notification procedure does not mean that the public safety commission has independently certified every technical component.

The resulting division of responsibilities can be summarized as follows.

This division provides a basis for deployment governance.

Its continuing effectiveness depends partly on whether the relevant actors receive adequate information when the system or its operating conditions change.

4. Governance After Deployment: Software, Supervision, and Incident Information

Initial conformity assessment and road-use notification establish important starting conditions. Robotic systems, however, do not necessarily remain in their original configurations.

Software may be updated, operating areas may expand, and organizations may increase the number of robots managed by one remote supervisor.

These developments raise a practical governance question: which actors must identify, evaluate, and communicate changes that could affect the assumptions underlying safe operation?

4.1 Remote supervision ratios: connecting assessment and operation

The number of robots managed by one remote supervisor provides a concrete example.

A system configured for one person supervising one robot presents different operational demands from a system in which one person monitors several autonomous robots simultaneously.

As the number increases, the supervisor may need to respond to multiple unexpected events, pedestrian interactions, communication failures, or requests for intervention.

It is important to distinguish three related functions.

Remote monitoring concerns observing the condition and behavior of robotic systems.

Remote operation concerns controlling their movement or other relevant functions from a distance.

Remote intervention concerns taking action when a robot requires human assistance, restriction, or emergency response.

These functions may be performed by the same person, but they are not technically identical.

The Robot Delivery Association's safety-standard development materials recognize the importance of evaluating simultaneous operation. Its framework addresses the number of robots managed by one remote operator and the conditions required for safe operation. ^[15]

The police notification framework considers the same issue from an administrative perspective.

The March 2023 administrative circular explains that descriptions of remote-operation arrangements may include the maximum number of vehicles one person can operate, staffing arrangements, training, and emergency-response procedures. ^[12]

The RDA's public list of assessed models also provides information about maximum simultaneous connection ratios, expressed as the number of remote monitors relative to the number of robots (遠隔監視者数：ロボット数).

For example, some listed configurations specify 1:4, while another specifies 2:4. ^[16]

These figures describe particular assessed configurations. They do not establish a universal statutory limit of four robots per supervisor.

In March 2025, NEDO reported that Panasonic Holdings had demonstrated the simultaneous operation of ten delivery robots under one remote monitoring and operating professional.

The experiment was conducted under a road-use permission procedure rather than the ordinary notification route. NEDO also indicated that further confirmation of the safety of simultaneous ten-robot operation would support practical deployment. ^[17]

The demonstration provides evidence of technical and operational progress under specific experimental conditions. It does not automatically authorize the same configuration for every robot or operator.

Together, these developments reveal a connection between three forms of evidence: private safety assessment, notified operating arrangements, and experimental performance.

When a supervision ratio changes, the relevant question is whether the new arrangement remains compatible with the assumptions underlying the robot's assessment and the operator's actual capabilities.

Relevant factors may include intervention frequency, operator workload, response time, communication reliability, and the ability to manage simultaneous incidents.

The governance challenge is maintaining consistency between the configuration assessed, the arrangements notified, and the system actually operated.

4.2 Software and operating changes after assessment

Similar questions arise when software or operating conditions change.

Japan's statutory notification items do not establish a separate requirement to identify every software version or navigation algorithm revision.

However, the police administrative interpretation explains that descriptions of remote-operation arrangements may include the programs used, control methods, emergency-stop notifications, and procedures for communication delays or interruptions. ^[12]

Software-related information may therefore form part of the documentation supporting a notification.

A change affecting legally specified notification matters may require an amended notification. A new version number is not, by itself, identified as a separate statutory notification category.

Private conformity assessment introduces another consideration.

As discussed above, the RDA's published guidance requires reassessment when the combination of an assessed robot and its remote-operation equipment changes. Changes to the operating environment or operating arrangements relative to those submitted for assessment require additional evaluation of the relevant differences. ^[13]

Different organizations may possess different information about these changes.

A manufacturer may know precisely what a software update modifies. An operator may understand whether the change affects actual routes, supervision arrangements, or system behavior. The assessment organization may know which configurations and conditions were covered by its evaluation. Public authorities may need information about changes affecting notified operating arrangements or traffic safety.

The governance issue is therefore not simply whether software updates are documented.

It is whether changes that materially affect safe operation are identified, assessed, and communicated to the actors responsible for making the relevant decisions.

4.3 Accident reporting and operational evidence

Accident and incident information provides another connection between initial assessment and continuing operation.

RDA materials presented during the government's public-private consultation process discussed arrangements for reporting accidents, legal violations, and related events to the association. ^[15]

The association has subsequently continued developing its conformity-assessment activities and publishes information about assessed robot models.

At METI's 11th Public-Private Council on Autonomous Delivery Robots in May 2026, the RDA presented an updated activity report. ^[18]

These materials provide evidence of the development of the assessment framework. However, conformity-assessment activity and operational safety outcomes are different measures.

The number of organizations receiving certificates cannot establish how many robots are operating, how far they have traveled, or how frequently incidents occur.

The publicly accessible materials examined for this article do not establish whether a comprehensive RDA accident-information collection mechanism has become fully operational or whether consolidated accident statistics have been published.

This is a limitation of the available evidence, rather than a finding that no incident information is being collected.

Evaluating real-world safety requires information about both incidents and operating exposure.

Incident counts become more meaningful when considered alongside distance traveled, operating hours, completed journeys, and the number of robots in service.

Near misses, emergency interventions, and accessibility complaints may also provide relevant evidence, depending on the environment.

The quality of these records matters because operational experience can reveal risks that were not fully captured during initial assessment.

4.4 The critical point of institutional coordination

The preceding examples identify the most important governance question in Japan's framework.

Initial technical assessment, road-use notification, and operational supervision are performed through different institutional mechanisms.

This division allows specialized organizations to carry out different responsibilities. Manufacturers understand technical changes, assessment organizations evaluate systems against defined criteria, operators manage actual deployment, and public authorities supervise road use.

Their responsibilities become interdependent when the assessed system or its operating conditions change.

A higher supervision ratio may alter operational risk without changing the robot's dimensions. A software update may affect navigation behavior without changing the name of the operator or the notified operating area. An incident may reveal information relevant to a manufacturer, an assessor, an operator, and a public authority simultaneously.

The most information-dependent point in this arrangement is therefore the transition from an initially assessed configuration to continuing operation under changing conditions.

This finding does not establish that information transfers routinely fail.

It identifies the institutional relationship whose effectiveness is particularly important to examine.

The practical question is whether existing assessment procedures, notification requirements, reporting powers, and operational records provide the necessary information to the appropriate actors when decisions about continued safe operation must be made.

5. European Regulation: A Comparative Perspective

European legislation provides a useful comparison because it addresses several risks arising from autonomous behavior and changes after deployment.

Regulation (EU) 2023/1230, the Machinery Regulation, will generally apply from 20 January 2027. It includes requirements relevant to autonomous machinery, including risks associated with fully or partially self-evolving behavior or logic. ^[19]

The EU Artificial Intelligence Act adds another regulatory dimension for systems falling within its defined scope.

Following Regulation (EU) 2026/1744, specified high-risk AI requirements are scheduled to apply from 2 December 2027 for systems covered by the relevant stand-alone high-risk provisions and from 2 August 2028 for qualifying high-risk AI systems associated with regulated products. Other provisions follow different application dates. ^[20]

The amendments also seek to improve coordination between AI obligations and existing product legislation, including the Machinery Regulation.

A particularly relevant development concerns product liability.

Directive (EU) 2024/2853, the revised Product Liability Directive, expressly includes software within its product framework and addresses circumstances involving software updates and upgrades.

Following a corrigendum published on 7 May 2026, the revised directive applies to products placed on the market or put into service after 8 December 2026, subject to the applicable national implementation arrangements. ^[21]

These provisions are relevant where changes to software or related components contribute to a defective product causing legally recognized damage.

However, product liability and operational authorization remain distinct.

Liability rules may address compensation for damage caused by defective products. They do not automatically determine where delivery robots may operate, how many robots one operator may supervise, or when changes to operating conditions require reassessment.

The European example therefore reinforces the importance of distinguishing technical safety, ongoing operation, liability, and public oversight.

It also shows that established legal systems are adapting to autonomous technologies, rather than leaving them entirely outside existing regulatory structures.

6. Lessons from the Cruise Case: Information and Regulatory Enforcement

A documented incident involving Cruise in California illustrates the importance of access to operational information after deployment.

On 2 October 2023, a pedestrian struck by a human-driven vehicle was thrown into the path of a Cruise autonomous vehicle.

The Cruise vehicle initially stopped but subsequently performed a maneuver that dragged the pedestrian approximately 20 feet.

On 24 October 2023, the California Department of Motor Vehicles suspended Cruise's autonomous vehicle deployment and driverless testing permits, citing safety concerns and its determination that Cruise had misrepresented information related to the safety of its autonomous technology. ^[22]

In September 2024, the National Highway Traffic Safety Administration announced a consent order concerning incomplete crash reports submitted by Cruise.

NHTSA stated that omitted post-crash details became apparent after the agency reviewed video requested from the company.

The order included a $1.5 million penalty and additional reporting and oversight obligations involving vehicle operations, software updates, and safety performance. ^[23]

The case illustrates two complementary aspects of governance.

First, existing regulatory institutions possessed substantial enforcement powers. The DMV could suspend autonomous operations, while NHTSA could impose penalties and additional reporting requirements.

Second, effective enforcement depended on obtaining complete and reliable information.

For complex automated systems, relevant evidence may be held primarily by manufacturers or operators. Regulators may therefore require mechanisms for obtaining, verifying, and interpreting technical information about actual system behavior.

A pre-deployment conformity assessment and a post-incident investigation serve different functions.

The former evaluates a system against defined requirements and assumptions. The latter examines what occurred during real-world operation, including the effects of changes or unexpected conditions.

Autonomous taxis and low-speed sidewalk delivery robots operate under different legal regimes and present different risk profiles.

Nevertheless, the Cruise case demonstrates a broader institutional principle: regulatory authority becomes more effective when responsible institutions can obtain and independently assess the evidence necessary for their decisions.

For Japan's delivery robot framework, this provides a reason to examine how technical assessment information, operational records, and incident evidence can support the statutory responsibilities of the relevant authorities.

7. Discussion and Conclusion: Governance Beyond Initial Assessment

Japan's delivery robot framework demonstrates that robot governance already operates through a division of institutional responsibilities.

Manufacturers and operators are responsible for their respective technical and operational activities. Private assessment organizations, including the Robot Delivery Association, can evaluate systems and issue conformity certificates. Public safety commissions and police authorities administer road-use notification, supervise traffic, and exercise statutory enforcement powers.

Third-party conformity certificates are an available form of technical evidence, while the law also accepts other qualifying objective documentation.

This arrangement separates technical assessment from public-road administration and provides an institutional structure for deployment.

The more demanding governance questions emerge when the conditions underlying those assessments and notifications change.

Software updates may alter navigation behavior. Operators may increase the number of robots supervised simultaneously. Operating environments may become more complex, and incidents may reveal risks not apparent during initial evaluation.

At that point, relevant information may be distributed across several organizations.

The manufacturer knows what changed technically. The operator knows how the system is actually being used. The assessment organization knows the scope and assumptions of the original evaluation. Public authorities possess legal powers to supervise road use and respond to relevant risks.

The critical institutional connection is therefore between initial assessment and continuing operation under changing conditions.

The RDA's supervision-related requirements, the police notification framework, and NEDO's 1:10 demonstration illustrate this relationship.

The development of incident-information arrangements further highlights the importance of learning from operational experience.

However, the effectiveness of these mechanisms cannot be determined from their formal existence alone. Evaluation must consider the accuracy of technical information, the timeliness of reporting, the relevance of operational evidence, and the ability of responsible institutions to intervene when necessary.

This does not mean that every change requires new legislation or an additional certification body. Existing mechanisms may already provide appropriate responses in many circumstances.

Additional requirements should be justified by identifiable problems and evaluated against their expected benefits, compliance costs, and effects on innovation.

The same evidentiary standard should apply to established arrangements: their effectiveness should be examined through operating experience rather than assumed from the existence of laws, certificates, or administrative procedures.

Japan's framework provides a useful basis for this evaluation because it makes several responsibilities identifiable while allowing technical assessments and operating practices to evolve.

The central finding of this article is that Japan has established separate mechanisms for robot safety assessment, public-road notification, and traffic oversight. The most information-dependent part of this arrangement is maintaining the connection between the configuration originally assessed, the conditions actually operated, and the evidence available when significant changes or incidents require review.

This relationship provides a concrete subject for further empirical research.

It also illustrates a broader principle for robot governance.

As autonomous machines become more capable and their operating conditions change, institutional effectiveness depends not only on assigning responsibilities at deployment, but on preserving the information necessary to exercise those responsibilities throughout operation.

References

1. Elon Musk and Walter Isaacson. Elon Musk: “Humans Won't Be in Control” When AI Becomes Smarter Than Us. The Builders, 20 October 2026; recorded 6 August 2026. Original podcast — Apple Podcasts
2. Bill Gates and Nilay Patel. Bill Gates says to prepare for 20 years of ‘tumultuous’ AI disruption. Decoder, The Verge, 22 October 2026. Interview and transcript — The Verge
3. Commonwealth Bank of Australia. ‘We try to think out loud’: OpenAI's Sam Altman on closing the gap between AI tech and how we're adopting it. May 2026. Official CBA report
4. Kaaren Morrissey / Australian Associated Press. Sam Altman thinks using AI in emails and Slack is ‘dehumanising’ – and revenue will ‘take a bit longer to figure out’. Startup Daily, 26 May 2026. Startup Daily report
5. International Organization for Standardization. ISO 12100:2010. Safety of machinery — General principles for design — Risk assessment and risk reduction. ISO official standard information
6. International Organization for Standardization. ISO 13482:2014. Robots and robotic devices — Safety requirements for personal care robots. ISO official standard information
7. International Organization for Standardization. ISO 3691-4:2023. Industrial trucks — Safety requirements and verification — Part 4: Driverless industrial trucks and their systems. ISO official standard information
8. International Organization for Standardization. ISO 31101:2023. Robotics — Application services provided by service robots — Safety management systems requirements. ISO official standard information
9. Robot Delivery Association / Ministry of Economy, Trade and Industry. IECにおける国際標準開発への参画. Materials presented at the 10th Public-Private Council on Autonomous Delivery Robots, 11 March 2025. METI official materials · Draft status — Danish Standards
10. Japan. Road Traffic Act (道路交通法), provisions concerning remotely operated small vehicles. Official legislation — e-Gov
11. Japan. Road Traffic Act Enforcement Regulations (道路交通法施行規則), including Articles 1-6, 1-7 and 5-4. Official regulations — e-Gov
12. Fukushima Prefectural Police. 遠隔操作型小型車の遠隔操作による通行の届出に関する解釈及び運用上の留意事項について（通達）. 23 March 2023. Administrative interpretation and implementation circular, not a statutory enactment. Official administrative circular
13. Robot Delivery Association. 安全基準適合審査について. Published assessment overview covering evaluation conditions, exclusions and reassessment requirements. RDA official assessment overview
14. Aichi Prefecture. 自動配送ロボットの公道走行によるラストワンマイル配送実証実験. 2 December 2025. Official Aichi Prefecture announcement
15. Robot Delivery Association / Ministry of Economy, Trade and Industry. 遠隔操作型小型車 RDA安全基準等策定状況. Materials presented at the 7th Public-Private Council on Autonomous Delivery Robots, 24 February 2023. METI official materials
16. Robot Delivery Association. 合格証を発行した機体. Public list of assessed robot models and maximum simultaneous connection ratios. RDA certified models
17. New Energy and Industrial Technology Development Organization (NEDO). 自動配送ロボットによる配送サービスの実用化への取り組み. 17 March 2025. NEDO official report
18. Ministry of Economy, Trade and Industry / Robot Delivery Association. 一般社団法人ロボットデリバリー協会 活動報告. Materials presented at the 11th Public-Private Council on Autonomous Delivery Robots, 19 May 2026. METI meeting materials
19. European Union. Regulation (EU) 2023/1230 on machinery and related products. Consolidated version incorporating amendments through July 2026. Machinery Regulation — EUR-Lex
20. European Union. Regulation (EU) 2026/1744 amending the Artificial Intelligence Act and related legislation. Amending Regulation — EUR-Lex
21. European Union. Directive (EU) 2024/2853 on liability for defective products; corrigendum published 7 May 2026. Product Liability Directive — EUR-Lex · 2026 Corrigendum — EUR-Lex
22. California Department of Motor Vehicles. DMV Statement on Cruise LLC Suspension. 24 October 2023. Official DMV statement
23. National Highway Traffic Safety Administration. NHTSA Announces Consent Order with Cruise After Company Failed to Fully Report Crash Involving Pedestrian. 30 September 2024. Official NHTSA enforcement announcement

← Back to article index