Responsibility is a process, not a promise.
Turn an AI responsibility statement into decisions someone can inspect: a named owner, evidence, limits, a response plan and a record of what changes.
Risk-management framework · Edited by PsyBorg · Updated 2026-11-16
Ask what the commitment changes
When an organization says it uses AI responsibly, ask for a concrete example of a decision that statement affects. Who decides whether a tool is suitable? What evidence supports that decision? Who can change or stop its use when a problem appears?
NIST's voluntary AI Risk Management Framework 1.0 organizes risk management around Govern, Map, Measure and Manage. Governance runs through the other functions, and the framework describes ongoing, iterative work rather than a fixed checklist. [1]
PSYBORG's practical recommendation is to keep a short decision record that another person can inspect and use during a handover.
Make a decision record useful
Start with the use of AI being considered, the people it could affect and the decision that needs an owner. Describe the actual workflow, including what happens before and after the AI produces its output.
PSYBORG suggests answering five questions in plain language.
- Purpose and boundary: what task may the system perform, and what uses fall outside the agreed scope?
- Responsibility: who can approve changes, investigate a problem and pause the workflow? Who covers an absence?
- Evidence: which version and operating conditions were tested, what was observed, and which important questions remain open?
- Response: what would trigger investigation, a pause or another action, and how can an affected person raise a problem?
- Reconsideration: when will the decision be reviewed, and which changes would bring that review forward?
Attach evidence that helps answer those questions. A link to an evaluation needs its relevant result and limits beside it. Assign open questions to someone who can investigate them, with a clear point for deciding what happens next.
A hypothetical example: a new document type
Imagine a small team using an AI tool to draft summaries of public meeting notes. Its initial evaluation covers typed documents in one language, with a person checking each draft before release.
Later, the team receives scanned handwritten notes and considers using the same workflow. The earlier evaluation does not answer how well this new input will work. In PSYBORG's suggested decision record, the team would identify the changed condition, assign someone to examine it and agree how the new documents will be handled while that work is incomplete.
This example is hypothetical. It illustrates how a change can create an evidence gap; it does not claim that handwritten documents always cause failure.
Plan for the person who inherits the work
NIST calls for documented responsibilities and communication lines (GOVERN 2.1), planned monitoring and review (GOVERN 1.5), and post-deployment processes covering matters such as feedback, incident response and change management (MANAGE 4.1). [1]
PSYBORG recommends checking whether a colleague could find the current decision, supporting evidence, unresolved issues and response route without relying on one person's memory. Include the relevant model or product version and who last reviewed the decision. Record why an action was taken so the next reviewer can assess the reasoning.
Keep this record proportionate to the consequences and complexity of the use. Documentation is useful when it supports decisions and follow-through; its existence alone cannot establish that controls work.
Sources and scope
[1] NIST, AI Risk Management Framework 1.0 Core, 2023. See the Core introduction, GOVERN 1.5 and 2.1, MAP 1.1, MEASURE 2.3 and MANAGE 4.1.
Read source (airc.nist.gov)
Source checked 16 November 2026. The page says a revised framework is in progress; this briefing refers to the named 2023 edition. The decision-record questions, handover recommendations and meeting-notes example are PSYBORG editorial analysis. They do not constitute NIST certification, a complete assessment or evidence that a particular deployment is safe.
Continue reading
NIST AI Risk Management Framework 1.0 explained
Read source (psyborg369.com)
A human in the loop needs a real way to act.
Read source (psyborg369.com)
READING ROOM ↗