Cognitive Sovereignty
The entitlement of a person whose case is decided by a court or tribunal that the reasoning, findings and authorities on which the decision rests were verified and are owned by a human adjudicator, so that no part of what is decided against them rests on machine output that no one checked.
When this decides something for you
A judge, a tribunal member or an arbitrator now has, on the same screen as the file, a system that will draft an order, summarise the pleadings and supply authorities for any proposition asked of it. The parties do not see what the system supplied. They see an order with citations in it. If one of those citations does not exist, the party who loses has been decided against on nothing, and unless someone reads the order against the reports, no one will know. The same is true of an advocate's written submission, a regulator's adjudication order and a tribunal's award. This page names what the person on the receiving end is entitled to.
The question
When a machine has taken part in producing a decision, what may the person decided against demand of the human who signed it?
The premise
A decision is an act of a person holding an office. The office carries the duty to read the record, find the facts, identify the law and apply it. A machine can be given any of these tasks. It cannot be given the office. Whatever the machine did, the person who signed remains the one who decided, and the person decided against remains entitled to a decision by that person.
How it works
Cognitive sovereignty locates the entitlement in the litigant, not in the judge or the advocate. The judge has a duty to decide on the record; the advocate has a duty to verify what is filed. Those duties are owed to the court and to the Bar Council. The litigant's entitlement is what gives the litigant standing to complain when the duty is not performed, and it fixes what the complaint is about: not that a machine was used, but that a human did not verify and does not own what the machine supplied.
The mechanism has three parts. Verification: every authority, finding and quotation in the decision was checked by the adjudicator against the primary record before it was relied on. Ownership: the adjudicator answers for each of them as if they had been typed by hand, because the signature makes them the adjudicator's. Visibility: the person decided against can point to a human who did both. Where all three hold, machine assistance in producing the decision is no concern of the litigant's. Where any one fails, the decision was not made by the person who signed it, and the entitlement is breached whatever the merits.
The entitlement is therefore indexed to verification and ownership, not to the presence of a human. A human was present at every stage of the case described below. That is what distinguishes it from the right not to be subject to a solely automated decision, which several data-protection regimes give and which is satisfied by human intervention. Intervention without verification is the failure this page names.
One case
Gummadi Usha Rani v. Sure Mallikarjuna Rao, SLP(C) No.7575 of 2026, Supreme Court of India, order dated 27 February 2026 (Narasimha and Aradhe JJ). In a suit for injunction, the trial court dismissed the defendants' objections to an Advocate Commissioner's report by an order dated 19 August 2025 and, in doing so, relied on four decisions cited as reported in the Supreme Court Cases. The defendants challenged the order on the ground that the four decisions did not exist. The High Court of Andhra Pradesh, in CRP No.2487 of 2025 decided on 21 January 2026, found that the citations were generated by artificial intelligence, recorded a word of caution, and dismissed the revision on the merits, affirming the trial court.
The Supreme Court issued notice, stayed further proceedings on the Commissioner's report, and said this of the process: the case "assumes considerable institutional concern, not because of the decision that was taken on the merits of the case, but about the process of adjudication and determination". It took cognisance of "the Trial Court deploying AI generated non-existing, fake or synthetic alleged judgments", stated that the matter "has a direct bearing on integrity of adjudicatory process", and declared that "a decision based on such non-existent and fake alleged judgments is not an error in the decision making. It would be a misconduct and legal consequence shall follow." Notice went to the Attorney General, the Solicitor General and the Bar Council of India, and a senior counsel was appointed to assist.
Three things in that order carry the definition. A human judge signed the order at every level; the entitlement was still breached. The breach was located in the process, not the outcome; the High Court had been prepared to let the outcome stand. And the consequence attached to the person holding the office, as misconduct, not to the tool.
The rule
A person decided against is entitled to a decision whose reasoning and authorities a human adjudicator verified and owns. Machine assistance does not reduce that entitlement, and human presence does not satisfy it.
The advocate's side and the constitutional argument
The first of the Five Doctrines in the author's published work states the same entitlement from the advocate's side: the duty to frame the question, verify the material and answer for what is filed, so that the client's case reaches the court as human judgment. The two are one doctrine seen from the two ends of the adjudicatory process.
Whether the entitlement is a right under Articles 14 and 21 of the Constitution of India is the argued position of the author's paper in the Journal on Human Rights: that a decision affecting life, liberty or property which rests on unverified machine output is not a decision by procedure established by law. No court has yet held this, and this page states it as argument, not as law.
Not to be confused with
The phrase is used elsewhere for other objects. Konigsberg (2026) uses it for an individual's authorship of their own AI-assisted thinking. Valente (2026) uses it for personal identity against engagement-optimised AI. Chow (2026) and Medina Osório (2026) use it for a State's or institution's autonomy over AI infrastructure. A 2025 preprint on AI memory uses it for the geopolitics of memory lock-in, and a 2026 neuroinformatics paper for neural-data security. The neurorights literature, from Ienca and Andorno (2017) through Chile's Law 21.383 of 2021 to UNESCO's 2025 Recommendation on the Ethics of Neurotechnology, protects the brain and mental states against neurotechnology. This page concerns none of those. It concerns the judgment applied to a person by a court or tribunal, and that person's entitlement that a human verified and owns it. No use located in English-language sources as of 25 September 2026 places the term in adjudication.
Provenance and method.
Where this was published
- Shivam Shukla
- AI for Indian Advocates: The Practitioner's Standard for Supervised Intelligence (2026)
- Part II
- the first of the Five Doctrines
- Shivam Shukla
- Cognitive Sovereignty as a Digital Human Right: Protecting Litigants from Unsupervised Artificial Intelligence in the Adjudicatory Process
- Journal on Human Rights
- Vol II (RMLNLU and UPHRC
- in press)
- Shivam Shukla
- From Human Oversight to Cognitive Sovereignty: A Process-Based Governance
- AI Law Politics
- Vol 2 No 1 (2026)
- DOI 10.5709/alp-02.01.2026-04
Changelog
- 2026-09-25, version 1. First published.