UNAM Orders 58,000 Applicants to Retake Entrance Exam
AI-assisted proctoring failed at scale in a high-stakes context - a concrete reminder that automated monitoring cannot substitute for end-to-end integrity design.
Key points
- UNAM's AI-assisted remote exam proctoring failed to prevent suspected widespread cheating affecting 58,000 applicants.
- Score distribution shifted sharply - 16.3% scored 100+ in 2026 versus a 3.5% historical average - triggering the review.
- The case is international with no direct APS angle; relevant as a cautionary AI assurance case study only.
Implications for Australian agencies
- Consider APS agencies using or evaluating AI-assisted proctoring or automated assessment tools could consider whether their monitoring stack includes distributional anomaly detection alongside individual-level flagging.
- Monitor Risk and assurance practitioners may want to monitor how this case is referenced in emerging AI assurance and automated decision-making guidance, given its scale and clear failure mode documentation.
Implications are AI-generated. Starting points, not advice — see methodology for how they're framed.
View original source
Copied.
"UNAM Orders 58,000 Applicants to Retake Entrance Exam"
Source: Let's Data Science – AI Governance
Published: 5 August 2026
URL: https://letsdatascience.com/news/unam-orders-58000-applicants-to-retake-exam-7cbf52fe
Mexico's National Autonomous University (UNAM) ordered approximately 58,000 undergraduate applicants to retake its entrance exam in person after an expert commission found evidence of suspected widespread cheating during the university's first fully remote admissions test. The online exam used Respondus LockDown Browser and AI-assisted webcam monitoring from Territorium, but anomalous score distributions - applicants scoring 100 or above rising from a 3.5% historical average to 16.3% - prompted suspension of enrolments and a formal investigation. The episode illustrates that browser lockdown and AI webcam monitoring constrain only what they can directly observe; external devices, leaked content, and cohort-level integrity remain separate control problems requiring audit design, escalation processes, and distributional monitoring.
Implications for Australian agencies:
- [Consider] APS agencies using or evaluating AI-assisted proctoring or automated assessment tools could consider whether their monitoring stack includes distributional anomaly detection alongside individual-level flagging.
- [Monitor] Risk and assurance practitioners may want to monitor how this case is referenced in emerging AI assurance and automated decision-making guidance, given its scale and clear failure mode documentation.
Retrieved from SIMS, 16 September 2026.