AI in Education AI Minute Newsroom 2026-08-19

A university system pointed a model at 14,000 syllabuses and asked it to find the forbidden topics

The New York Times reported on 18 August that the Texas Tech University System, which teaches about 64,000 students across undergraduate, medical, law, nursing, pharmacy, dental and veterinary programmes, built an AI tool to search course materials for books and readings touching sexual orientation and gender identity. Its regents reviewed more than 14,000 courses offered in spring 2026 and recommended changes to fewer than 60; degree programmes centred on sexual orientation or gender identity were shut down in April. Professors told the newspaper that the tool scanned syllabuses and reading lists before any human looked at them, and described a system in which they now read between the lines and pre-emptively remove material rather than risk a report. The chancellor, Brandon Creighton, wrote the state law he is now enforcing: as a senator he authored SB 37, which handed governor-appointed regents authority over curriculum. In July the American Association of University Professors, Texas AAUP-AFT, Lambda Legal and the NAACP Legal Defense Fund sued him and the board, listing Plato's Republic pulled from an introductory philosophy course, Dred Scott v. Sandford restricted in law teaching, and medical instruction on treating racial and sexual minorities barred at the health science campuses.

Why it mattersCurriculum review is old; the AI here is not doing anything a committee could not do, only doing it to 14,000 courses at once instead of a handful. That is the whole change. When the cost of screening everything falls to nearly zero, the natural scope of a rule stops being what administrators have time to check and becomes everything, and the pressure lands on instructors long before any human reads a word. Note also what the numbers say: fewer than 60 courses were actually changed. The effect described by the professors is not the flagging, it is the anticipation of it.

✓ Verified · 4 sources

Read in the app — free, in 9 languages

A university system pointed a model at 14,000 syllabuses and asked it to find the forbidden topics

Related stories