The article argues that AI-mediated governance creates a political problem, not only a technical or regulatory one. When systems make or shape decisions through processes that officials and citizens cannot meaningfully interpret, responsibility can become dispersed and public consent weakened.

Drawing on Locke’s view of government as a trust granted by free and equal people, the article proposes three conditions for legitimate algorithmic authority: intelligibility, fiduciary accountability and contestability. It discusses technological civic literacy, participatory oversight, audits, explanation rights and constitutional safeguards as ways institutions could support those conditions, while stressing that these are not a technical blueprint.