The Hidden Stakeholders of Artificial Intelligence - How to Build AI Projects That Don’t Harm People #991711

When AI systems cause harm, who is responsible?
In many AI projects, the answer is unclear. Responsibility is shared across designers, project owners, operators, and vendors—and the people most affected are often the ones who never had a seat at the table.

The Hidden Stakeholders of Artificial Intelligence gives those people a name: passive stakeholders. They are the job applicants screened by an algorithm, the loan applicants denied by a model, and the communities flagged by facial recognition systems. They are affected by the system but have no power to influence the project that created it.

Drawing on more than twenty years of AI project experience and five cases drawn from regulatory actions and court proceedings—in retail surveillance, financial trading, auto lending, copyright and training data, and AI-generated legal citations—Gloria J. Miller introduces a structured framework that project teams can apply immediately.

The Four Tests of power, legitimacy, urgency, and harm help identify who is at risk before systems are deployed. A stakeholder workshop, harm taxonomy, and lifecycle-weighted scoring model translate the framework into practical tools for real projects.

AI projects are not complete when the system is delivered.
They are complete only when the risks to all affected stakeholders have been considered.