Sao Paulo's ChatGPT Lesson-Content Mandate -- and the Peer-Reviewed Backlash Over Quality and Equity

In April 2024, the government of São Paulo state — Brazil's largest public school system — announced that Seduc-SP, its state education department, would use ChatGPT to draft the first version of digital lesson presentations for grades 6-9 of elementary school and all three years of high school.

Under the announced workflow, the AI would generate a presentation from themes supplied by curriculum staff at a targeted pace of about three presentations every two days, after which a classroom teacher would review and correct the content before it reached students; officials described it as a pilot project. The plan drew immediate opposition from teachers' unions — Apeoesp called it an outrageous attack on educators' rights and Sinpeem said it undermined the teaching and learning process — and a state legislator filed a formal complaint with the Public Prosecutor's Office.

In July 2024, a peer-reviewed critique by four Brazilian researchers (Matos, Girotto Junior, de Medeiros Arnt and Lippi), published in the Cell Press journal Patterns, argued the proposal was unready: generative AI outputs carry biases and factual errors without reliable source citations, and reviewing AI drafts thoroughly takes real time that the plan's training-free rollout did not budget for.

The critique also raised a regional-equity objection: Brazil's North, Center-West and Northeast regions hold 82% of national territory and 43% of the population but have far weaker teacher-qualification rates than the Southeast (e.g. Amazonas has roughly 80% of teachers considered inadequately trained for their subject, against 60-80% adequately trained in São Paulo), warning that a one-size-fits-all AI content pipeline could deepen rather than close that gap. As of the critique's publication, Seduc-SP had not announced new teacher-training measures to accompany the rollout.