AI Disclosure Must Lead to Answerable Human Judgment

A court can disclose that artificial intelligence assisted a translation while leaving litigants unable to know whether meaning changed. A lecturer can require students to tick an AI-use box without explaining which assistance is permitted. A church can label an image as synthetic while nobody takes responsibility for the testimony it appears to depict. In each case, something has been disclosed, but little has become answerable.

Transparency is often presented as the easy part of AI governance. Put a label on the output, add a statement to the policy or ask the user to declare the tool. Yet a bare label can become administrative theatre. It may reveal neither what the system did nor what a responsible person checked. It can also demand so much data that learners, workers and vulnerable people are placed under unnecessary surveillance.

The better standard is answerable disclosure. When AI materially shapes evidence, meaning, communication or a consequential decision, the institution should reveal the role it played, identify the accountable human, preserve enough evidence for review and provide a route to question and remedy. Disclosure should make responsibility visible; it should not become a substitute for responsibility.

On 25 September 2026, UNESCO reported on a four-day dialogue with the African Court on Human and Peoples' Rights held during the Court's 82nd Ordinary Session. Seventy-two judges, legal officers and staff members…