Robotics evidence guide · automated editorial preparation, not expert review
Distinguish requesting a handoff from completing one
A handoff has separate request, acknowledgement and observed-transfer stages. Software acknowledgement alone does not establish that a person received the object or responsibility.
Evidence and interpretation
Maha proposes an evidence description for handoffs, using the AI risk framework only as broad lifecycle context. First specify what transfers: an object, a task, or supervisory responsibility. These are different events. Record how each endpoint is observed, which party can decline, and what remains uncertain. A completed physical transfer does not itself show comfort, accessibility or agreement; those require their own appropriately designed observations.
Proposed evidence workflow
- Name the object or responsibility and define an observable completion condition.
- Record request, acknowledgement, refusal and observed outcome separately.
- Preserve ambiguous or interrupted transfers rather than inferring completion from the last command.
Worked illustration
A robot reports that it opened its gripper after an acknowledgement. Without an observation of receipt, the proposed packet records commanded release, not successful delivery to a person. The missing evidence must not be supplied by a confident narrative.
Limits
No safe handover trajectory, force limit or human-subject validation is provided. The example does not operate a robot.
Sources and review
Locator: Overview of the AI RMF
The framework is voluntary and addresses risk considerations across AI design, development, use and evaluation.
Boundary: This overview does not determine applicable law, certify robots or validate Maha workflows.
Inspection and reuse basis
Selected sections inspected 2026-09-15. Original explanation and link only; no framework text redistributed. This is not independent expert review of this article.