The Palo Alto Lawsuit: A Detector Score Becomes a Civil Rights Case

The essay was on The Crucible. The lawsuit is about civil rights.

In October 2025, a sophomore at Palo Alto High School submitted an English essay on Arthur Miller's play to his teacher, Sarah Bartlett. Turnitin's AI-writing detector reported that 76 percent of the text was likely AI-generated. Bartlett applied the school's policy for suspected AI use: the student had to rewrite the essay in class, under supervision.

His family disagreed, and did something most accused students never do. They assembled roughly 1,200 pages of evidence — drafts, handwritten notes, and the document's full revision history — showing the essay built up line by line over days. They asked the district for a neutral grade by 6 March. It wasn't granted. The student sat the in-class rewrite, scored a D, and his final grade dropped to a C.

In May 2026, the boy's father, Takashi Kato, filed a federal lawsuit against Palo Alto Unified School District seeking $150 million, according to reporting by Palo Alto Online and the SF Standard. The complaint alleges discrimination, retaliation, denial of due process and improper grading — and claims his son, a multilingual Asian student, was treated differently from peers, citing at least one other Asian student in the same class who was flagged, rewrote in class, and also received a D. The district has denied all claims.

Why this case is different

Most disputes over AI detection so far have argued that the tools are inaccurate — that a Turnitin percentage is a probability score, not proof, and shouldn't be treated as one. That argument has already reshaped exam-board guidance in England, where JCQ has been explicit that a detection score alone cannot support a malpractice finding.

The Palo Alto complaint goes further. It doesn't just say the detector was wrong. It says the way the school acted on the detector's output was discriminatory and procedurally unfair — no clear appeal route, no independent review of the family's evidence, and a disparate outcome for students from a particular demographic group. That reframes the risk for schools. An inaccurate tool is a pedagogical problem. A policy that applies an inaccurate tool inconsistently across student groups is a legal one.

The evidentiary weakness of AI detectors has been known for years. What's new is a court being asked whether a school's response to that weak evidence violated a student's civil rights.

The underlying accuracy problem isn't new. A widely cited 2023 Stanford study found that AI detectors misclassified more than half of TOEFL essays written by non-native English speakers as AI-generated, with an average false positive rate above 60 percent — because simpler, more formulaic sentence construction common among second-language writers resembles patterns the detectors associate with machine text. That finding predates this case by three years, but it sits squarely inside its central claim: that detection tools may not fail randomly, they may fail in a patterned way that lands hardest on specific groups of students.

Other cases are testing similar ground. A federal judge ruled in February 2026 that an AI-plagiarism finding against Adelphi University student Orion Newby was "without merit" and ordered it removed from his record. FutureEd and legal trackers now count at least half a dozen active US lawsuits over AI-detection decisions, with courts showing more scepticism when a school's case rests on a detector score alone, and more deference when the flag is backed by independent evidence.

What UK schools should take from a US case

England isn't the United States, and Title VI-style discrimination claims don't map directly onto UK law. But the operational lesson travels well. If a policy uses an AI detector as a trigger for disciplinary consequences, three questions now matter more than they did a year ago: is the process the same for every student regardless of background or language profile; is there a genuine route to challenge the finding with evidence, not just resubmit under supervision; and is the decision documented well enough to survive a legal challenge, not just a parent complaint.

Schools that already treat a detection score as one input among several — alongside drafts, in-class writing samples, and teacher judgement — are in a stronger position than those using it as the trigger for a set consequence. The Palo Alto case suggests that "the software flagged it" is no longer a defensible endpoint for a decision that affects a student's grade.

What to do

Audit any policy that lets an AI-detection score alone trigger a grade penalty or disciplinary process, and separate the flag from the consequence. Build in a documented, evidence-based appeal step before any grade is changed, not after. Ask whether your assessment policy asks for process evidence — drafts, edit history, in-class writing — as standard, so it exists before a dispute rather than being assembled defensively afterwards.

What to watch

The Palo Alto case is at an early stage and the district is contesting it; a ruling, if it comes, could set persuasive precedent for how US schools are expected to handle detector-based accusations. Watch also for whether UK unions or the Department for Education issue equivalent guidance on due process and equality duties around AI-detection use, given that the accuracy concerns raised by the Stanford research apply just as much to EAL pupils in English classrooms as to multilingual students in California.