What a Court Needed Beyond a 100% AI Score

The short answer. A New York court annulled an academic integrity finding against a first-year student that rested on a Turnitin "AI-generated score of 100%", and ordered the university to expunge his record. The court did not decide whether the essay was written by AI, and it said nothing about whether detectors work. It found the university had not considered the student's evidence, had not given him the advisor its own rules promised, and had let the same official decide the case and then hear the appeal. The score was never the problem. What the university built on it was.

If you run a department, the case you are most likely to hear about this term is Matter of Newby v Adelphi University. It is being passed around as "the court that threw out an AI detector". Read the judgment and it is a narrower, more useful thing than that. It is a list of what a misconduct process needs to have in place before a detector result is allowed to carry any weight at all, written by a judge who was not asked to rate the detector and did not try to.

What happened

The facts are from the court's own opinion, decided in the Supreme Court of New York, Nassau County, on 28 January 2026, and reported as 2026 NY Slip Op 26021.

In autumn 2024 a first-year student submitted an essay for a course called World Civilization. He was enrolled in the university's Bridges programme, which the opinion describes as supporting students "who self disclose with nonverbal and neurosocial disorders, including Autism Spectrum Disorder", and had worked on the essay with a tutor from that programme. His professor ran the essay through Turnitin. In the court's words, the professor "employed 'Turnitin', an AI detection service which produced an 'AI-generated score of 100%'", and concluded the essay "had 'been produced by artificial intelligence'".

The student objected in writing. He "categorically denied using AI to write the Essay", and his parents ran the same text through two other detectors, Grammarly's and ZeroGPT, which "purportedly demonstrated that the Essay had 'a 0% chance of [being] AI written'". The university's academic integrity officer found him responsible anyway and required him to attend a plagiarism workshop. He appealed. The same officer denied the appeal.

Then something unusual happened. On 4 December 2024 the professor emailed the student to say he had been "under the impression" that the integrity office "was going to take your file into consideration, make their own decision as to whether you were responsible for an academic integrity violation", and that "it seems like their system defaults you as responsible unless you contest the claim again". He suggested the student write back "simply reminding him that you disagree, that your [E]ssay was not written with AI".

The student went to court.

What the court decided, and what it did not

New York courts review a private university's disciplinary decisions on a deliberately narrow standard. The opinion states it: review "is limited to a determination as to whether the school acted arbitrarily and capriciously, or whether it substantially complied with its own rules and regulations". The judge was not deciding whether the student used AI. She was deciding whether the university had followed its own process and reached its decision on a rational basis.

On both counts it had not. Three findings do the work.

The student's evidence was never considered. The court found the integrity officer "failed to even consider the Petitioner's evidence" in denying the appeal. Two detector results pointing the other way, and a denial in writing, went unanswered.

The university broke its own rules. Its Student Bill of Rights says every student may "be accompanied by an advisor of choice who may assist and advise" them "throughout the University disciplinary proceedings including all meetings and hearings". The court found the student "was not afforded the opportunity to confer with an advisor of his choice regarding the Violation".

The appeal went to the person who made the decision. The officer who issued the finding "was the identical person charged with the responsibility of entertaining the appeal", which the court said thwarts "a student's right to an avenue of meaningful 'appeal'".

And the professor's email mattered. The court held that "as the Oelze Email, at a minimum, undermines both the strength and substance of the plagiarism claims alleged in the Violation, this Court finds said Violation and the Denial directly resulting therefrom to be without valid basis and devoid of reason". The order: the violation and the appeal denial "are annulled", and the university "is directed to expunge the Petitioner's academic record with respect to the Violation and any sanction imposed in connection therewith is rescinded".

Notice what is absent. There is no finding that Turnitin was wrong. There is no finding that detectors are unreliable. The court did not need either. A 100% score, a written denial, two contrary results and a professor's own doubt, put through a process with no advisor, no consideration of the response and no independent appeal, was enough on its own. This is one decision by one state court about one university's process. It binds nobody in Britain. It is still the clearest worked example yet of a detector result being asked to carry a case, and the case collapsing under it.

The UK already has the same instruction in writing

Nothing the court required is foreign to a British school. It is already in the documents most staffrooms have not read.

JCQ's guidance on AI use in assessments, which we went through line by line in what JCQ actually says about AI detection in coursework, treats a detector result as one piece of evidence inside a holistic judgement of whether work is authentic, a judgement made by the teacher, never by the tool.

The University of Manchester's AI guidelines go further: "Tools to detect AI-generated content are unreliable and biased and cannot be relied on to identify academic malpractice in summative assessment. Output from such tools cannot currently be used as evidence of malpractice." Its malpractice guidance points staff to the wider evidence instead, prompts left in submitted work, fabricated references, discussion with the student, vivas.

The Dubai Accord, launched 27 August, says "no automated detection output or probabilistic indicator should be treated as conclusive evidence of misconduct", and asks institutions to give students "notice of concerns, access to the substance of the evidence, a meaningful opportunity to respond and a route to independent review or appeal". Three of those four are things the New York court found missing.

And the University of British Columbia now requires every syllabus to state what AI use is permitted before the term starts, so the first question in any case, was this allowed, has an answer in writing.

Four institutions in one month, plus a court, saying the same thing from different directions. The score starts a conversation. It does not close a case.

What a school should have before it makes a finding

Take the New York case as a checklist of what the university did not have. Before a detector result is allowed to carry any weight:

• A written rule the student could have read. What AI use was permitted on this task, stated before submission, not reconstructed afterwards.

• The student's account, considered and answered. A denial, a draft history, a tutor's involvement, a contrary detector result. Each one written down and responded to in writing, not defaulted past.

• Process evidence, not just output evidence. Version history, planning notes, a conversation about the work, a short viva. This is what JCQ and Manchester both point to, and it is what a detector cannot produce.

• Someone in the student's corner. An adult of their choosing in the room, if your own policy promises it. If it promises it and does not deliver it, that alone was enough for this court.

• An appeal to a different person. The finding and the appeal cannot sit with the same official. This was the simplest failure in the case, and the most avoidable.

Our JCQ checklist for the new term covers the first three in one printable page, and the school AI-use policy template covers the fourth and fifth. If a student in your school has already been flagged, our guide for wrongly flagged students sets out what they can reasonably ask for, which is much the same list from the other side.

Where that leaves a detection vendor

We sell an AI detector, so the fair reading of this case is against us as much as anyone. A tool that reported 100% on an essay a student wrote with a tutor is the failure every teacher fears, and no detector, ours included, can promise it will not happen. Our own published error rates say as much: our current model wrongly flagged 4 of 400 real student essays, catches about 24 in 100 essays copied straight from a chatbot, and caught none of 70 submissions that were half AI and half the student's own writing. Those numbers are on our methodology page with the cases we miss left in.

Every result we return is labelled a screening signal. After this case, that label reads less like caution and more like the only honest description of what any detector output is. The court in Nassau County did not need to say detectors are unreliable. It only needed a university to treat one as if it were reliable, and the process fell over.

Who wrote this, and what we sell. Is It AI is an AI-writing detector for teachers. The quotations above are from the court's published opinion and from the institutions' own pages, linked in each case. We have no connection to the student, the university or Turnitin.

Sources

• Supreme Court of New York, Nassau County, Matter of Newby v Adelphi University, 2026 NY Slip Op 26021, decided 28 January 2026 (Marber, J.). Also reproduced by FindLaw. All quotations from the case are from the opinion.

• University of Manchester, University AI Guidelines.

• Joint Council for Qualifications, AI Use in Assessments, revision two, 30 April 2025.

• European Network for Academic Integrity, Dubai Accord on Academic Integrity in the Age of Artificial Intelligence, 27 August 2026.

• Inside Higher Ed, Adelphi Student Wins AI Plagiarism Lawsuit, 11 February 2026, for the university's statement that it does not comment on litigation.