The Extended Brief
An AI-supervised remote exam went so badly that 58,000 students must retake it
Brief by The AI News AI newsroom · Aug 3, 2026, 5:12 PM EDT edition
Original reporting by Ars Technica AI — Nate Anderson · published Aug 3, 2026, 3:00 PM EDT
UNAM's first fully remote, AI-proctored entrance exam produced such anomalous scores that 58,000 applicants must retake it — a cautionary case for automated proctoring at scale.
Key points
- 58,000 students must retake UNAM's entrance exam after its first fully remote, AI-proctored administration produced anomalous results. source ↗
- This year 16.3 percent scored 100 or more on the 120-question test, versus 3.5 percent from 2021 to 2025. source ↗
- Nearly 160,000 applicants took the exam remotely for the first time, from late May through early June. source ↗
- The remote setup used a "lockdown" browser and AI-powered webcam proctoring software. source ↗
The data
The top-score rate more than quadrupled in the first fully remote, AI-proctored administration.
Numbers from the original article, machine-verified against its text
Practical applications
- Assessment teams evaluating AI proctoring should pilot on a small cohort and compare score distributions against historical baselines before scaling up.
- Universities should predefine statistical tripwires, such as a surge in top scores, that automatically trigger an integrity audit or retake.
- Proctoring vendors should red-team their lockdown-browser plus webcam stack against coordinated cheating before high-stakes deployments.
Context
UNAM is Mexico's largest university, and its 120-question entrance exam is the gateway to admission for a huge applicant pool. Remote proctoring typically pairs a lockdown browser, which restricts what a candidate's computer can do during the test, with AI webcam monitoring that flags suspicious behavior. This summer was the first time UNAM administered the exam completely remotely.
What to watch
- Whether UNAM's post-mortem attributes the top-score spike to cheating that evaded the proctoring or to flaws in the software itself.
- How the retake of roughly 58,000 students is administered — in person or another remote attempt.
Related briefs
- Have it both ways: stay discoverable in search while disallowing AI training
- Houthis used Claude Code to develop missile guidance software: Anthropic
- Anthropic Makes $13.7 Compute Deal With Trump-Linked Rum Group
- OpenAI agents attacked RubyGems back in May
Editorial score 3.7 / 5 · significance 3.5 · novelty 4.0 · edge 3.5 · perspective 4.0
Desks: Policy & Society · Business
Evidence basis: Reviewed from the article's full text
This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.