Wingward.ai — Navigate AI with clarity and purpose.
Menu

The Wingward Perspective

Accountability at the Center of Responsible AI

An organization’s commitment to responsible AI becomes meaningful when people can explain its decisions, evaluate its effects, and correct problems. That requires clear responsibility throughout the system’s life.

Brian Peterson
Brian PetersonWingward · September 10, 2026 · 4 min read
A PRACTICAL LEADERSHIP MODEL

The Responsible AI Wheel

Six commitments connected by accountability

The Responsible AI Wheel Accountability connects six ethical commitments: fairness and inclusion; human agency and oversight; privacy and data governance; social and environmental well-being; safety and robustness; transparency and explainability. The outer loop represents governance throughout the AI lifecycle. Segment size and numbering indicate neither importance nor sequence. 01Fairness &inclusion 02Human agency& oversight 03Privacy &data governance 04Social & environmentalwell-being 05Safety &robustness 06Transparency& explainability AT THE CENTERAccountabilityResponsibilityand auditability CONTINUOUS GOVERNANCE
  1. 01Fairness & inclusion
  2. 02Human agency & oversight
  3. 03Privacy & data governance
  4. 04Social & environmental well-being
  5. 05Safety & robustness
  6. 06Transparency & explainability
Accountability connects every commitment. Governance continues from initial design through retirement.Diagram arrangement by Wingward Consulting.

The Responsible AI Wheel offers a simple way to think about this challenge. It places accountability at the center, connects it to six ethical commitments, and surrounds the whole model with a cycle of continuing governance.

Papagiannidis and colleagues frame accountability in terms of responsibility and auditability. Responsibility means assigning people, roles, or departments to oversee AI systems and the decisions made with them. Auditability means enabling assessment of a system’s algorithms, data, and design processes. Together, they connect ownership with evidence that others can examine. These responsibilities extend across leadership, technical teams, operational staff, and external providers.

Six commitments surround that center:

01

Fairness and inclusion

Examine who benefits, who may be disadvantaged, and whether people can access and use the system.

02

Human agency and oversight

Preserve meaningful human choice and the ability to review, challenge, or intervene.

03

Privacy and data governance

Protect information and ensure that data are appropriate, sufficiently accurate, and used responsibly.

04

Social and environmental well-being

Consider consequences for communities, working lives, and the environment.

05

Safety and robustness

Evaluate whether the system performs reliably, withstands disruption, and avoids foreseeable harm.

06

Transparency and explainability

Make AI use visible and provide understandable information about its operation, limitations, and outputs.

These commitments draw on the principles synthesized by Papagiannidis and colleagues. Their positions around the wheel emphasize that each deserves attention. Decisions about one can affect the others.

The outer loop represents continuous governance. Responsibilities begin when an organization considers an AI use case and continue through acquisition, testing, deployment, monitoring, improvement, and retirement. Laws, ethical expectations, and stakeholder needs inform that work. This reflects Mäntymäki and colleagues’ hourglass model, which connects external expectations, organizational governance, and the operation of individual AI systems.

Accountability in practice: Rite Aid

Rite Aid’s experience shows what is at stake. In a 2023 complaint, the Federal Trade Commission alleged that the retailer’s facial-recognition system generated thousands of false matches. According to the complaint, employees acted on incorrect alerts by confronting customers, removing them from stores, or calling police. The FTC alleged failures to assess accuracy, train and oversee employees, and monitor errors after deployment. FTC complaint

A 2024 settlement imposed a five-year prohibition on facial recognition for security or surveillance. It also required designated responsibility, documented risk assessments, and ongoing monitoring for permitted biometric security systems. Rite Aid neither admitted nor denied the allegations. Settlement order

This wheel is a proposed synthesis for discussion and planning. The Rite Aid case illustrates its central concern: ethical commitments need to be connected to people with clear responsibilities and evidence that others can examine.

Those responsibilities must include the authority and practical ability to intervene when a system produces results inconsistent with its intended use. Organizations should identify who can override an AI-supported decision or deactivate a system, who leads the response to an incident, and who reviews appeals from people affected by its use. The NIST AI Risk Management Framework connects these actions with assigned responsibilities, monitoring, and documented responses (MANAGE 2.4, 4.1, and 4.3). Recording the reasons for interventions and their outcomes supports auditability and helps the organization determine whether its safeguards are working.

Resources