Skip to content

Responsibility

Physical capability must advance alongside accountability.

Humanoid robotics should be developed with clear operating boundaries, appropriate intervention, transparent limitations, privacy-conscious systems, and evaluation suited to real human environments.

Framework

Six principles for responsible development.

  • Human authority

    People remain responsible for permission, supervision, interruption, and appropriate use.

  • Bounded behavior

    Systems operate within supported environments, activities, and technical limits.

  • Proportionate supervision

    Supervision requirements reflect the task, environment, uncertainty, and potential consequence.

  • Privacy-conscious design

    Data practices should be limited, explainable, controlled, and appropriate to the feature being provided.

  • Secure operation

    Access, updates, identity, connected infrastructure, and system integrity require deliberate protection and ongoing maintenance.

  • Transparent communication

    Capabilities, limitations, connected behavior, and unresolved risks should be described accurately.

Supervision Framework

Four levels of expected human involvement.

Every supported activity should carry a clear supervision expectation. The distinction between supported, recommended, required, and not intended must be visible rather than hidden.

  • 01

    Supported

    Designed for the activity when the environment, setup, and operating conditions are appropriate.

  • 02

    Recommended

    Human supervision is recommended because the activity or environment may require judgment or intervention.

  • 03

    Required

    Active human supervision is required throughout the activity.

  • 04

    Not intended

    The system is not designed or approved for the activity.

Privacy

The physical environment can reveal sensitive information.

Humanoid systems may process information about spaces, objects, activity, schedules, and connected accounts. Responsible architecture should limit collection, define purpose, distinguish local and connected processing, protect access, and provide meaningful controls.

  • Collect only what is needed

    Data collection should be scoped to the feature being provided.

  • Explain why it is needed

    Purpose should be described in language a customer can understand.

  • Protect access and transmission

    Authorization, encryption, and integrity controls should apply throughout.

  • Provide retention and permission controls

    People should be able to manage what is kept and what remains active.

Not all information can remain local. Where connected processing is required, it should be described clearly, not implied to be entirely on-device.

Security

Security is an ongoing operating discipline.

No connected system should be described as invulnerable. Responsible security includes access control, protected updates, dependency management, monitoring, incident handling, recovery planning, and continuous review.

Boundaries

Clear boundaries support safer development.

Humanoid systems should not be presented as medical devices, emergency responders, childcare providers, security guards, or replacements for professional judgment unless separately designed, evaluated, and authorized for those purposes.

Responsible Expansion

Expand only when evidence supports expansion.

New capabilities should be introduced through defined environments, documented controls, appropriate evaluation, monitored operation, and clear communication about what has changed.