Yusril Says AI Accuracy Gives No Authority, Sets Out Five Constitutional Tests for Government Use
Yusril Ihza Mahendra, Indonesia’s coordinating minister for law, human rights, immigration and corrections, has argued that the central question in governing artificial intelligence is not how to give laws to machines but how to keep the people and institutions using them under the law, Jakarta Satu reported. He made the case in a speech to the 2026 International Conference on Democracy in National Resilience, organised by Universitas Sebelas Maret on managing AI for democracy, which he posted on his YouTube channel on Sunday, October 11.
Yusril noted that AI is not recognised as a legal subject under Indonesian law. When AI is used in actions with legal consequences, he said, the chain of responsibility cannot stop at the machine; it must lead back to whoever designed, supplied, selected, operated and, above all, acted on the system’s output. However accurate a model is, he said, accuracy does not confer a right to govern, because authority comes from law. He opened with scenarios that may soon be common: a person refused a service after being rated high risk, or a migrant pulled aside because a model flagged a suspicious pattern. The state, he said, cannot answer such people simply by saying the system decided.
What it means in practice
Yusril listed five constitutional problems to keep testing whenever AI enters public power. On authority, he cited Article 8 of Law No. 30 of 2014 on Government Administration, which requires decisions to be made by an authorised official, even when a human signature masks a machine’s judgement. On procedural fairness, citizens need to know whether AI was used, which factors mattered, who is responsible and where to object, with agencies supplying reasons, documentation and remedies. On equality, he warned that historical data can carry old biases and neutral-looking variables can stand in for identity. On privacy, he cited a Constitutional Court ruling on personal data and said not everything technically possible is lawful for the state. On the information space, he favoured proportionate duties on the provenance and labelling of synthetic content over limits on legitimate criticism.
The speech adds a legal frame to the government’s regulatory build-out, which includes draft presidential rules on an AI roadmap and ethics, the AI ethics and safety regulation now in final review and the One Data Indonesia law covering AI. It also sits alongside Komdigi’s preference for sector-by-sector rules over an AI slowdown.
What to watch: whether these principles appear in the final presidential regulations, and whether agencies that use AI start publishing explanations and objection channels for citizens.