AI proctoring can turn gaze, movement, sound, browser activity or other signals into suspicion of prohibited test behaviour. Under Annex III, systems used to monitor and detect prohibited behaviour of students during tests can be high-risk. The governance burden is preserving the evidence, human review and appeal path between an automated flag and any academic consequence.
The Regulation classifies specified AI used to monitor and detect prohibited behaviour of students during tests as high-risk. Article 5 separately prohibits emotion inference in education institutions except for medical or safety reasons. Current Commission guidance places Annex III education high-risk rules on the 2 December 2027 timeline.
Start with the actual system, role, use case and evidence boundary. The same regulation can produce different obligations for different actors and systems.
Annex III includes AI intended to monitor and detect prohibited behaviour of students during tests as a high-risk education use. Remote or in-person proctoring can therefore enter the high-risk route when the system materially flags or influences academic misconduct decisions.
Article 5 separately prohibits AI used to infer emotions of natural persons in education institutions, except where the system is intended for medical or safety reasons. A proctoring system should be screened for emotion-recognition features before any high-risk compliance analysis.
High-risk AI requires effective human oversight. A suspicious-behaviour flag should not silently become an academic penalty without a reviewable path that allows a competent human to understand limitations, examine evidence and disregard or reverse the output where appropriate.
Preserve the system and model version, monitored signals, trigger or score, captured evidence, human reviewer, override, notice, academic-integrity decision, appeal or challenge, logs, limitations and the basis for the final outcome.
The goal is not merely to reach an answer. It is to preserve what facts, evidence, scope and limitations supported that answer at that time.
Separate identity verification, browser or device monitoring, gaze or movement analysis, audio monitoring, object detection, behaviour scoring, anomaly detection and misconduct classification. The real function determines the legal route.
If the system infers emotions in an education setting, assess Article 5 first. Prohibited emotion inference cannot be made acceptable merely through better notices, logging or human review.
Connect the detected event, underlying evidence, system confidence or score, reviewer, decision authority, override, notice, appeal route and final academic outcome so a false positive can be challenged and corrected.
A new model, changed detection threshold, new sensor, different exam format, broader monitoring scope, removed human checkpoint or new misconduct rule can make the prior classification and evidence stale.
Choose the smallest operating tier that fits the portfolio today. Upgrade when system count, team size or governance scope actually requires it.
Keep a living system-level evidence record with obligations, gaps, versions and revalidation state.
START EVIDENCE PASSPORT →Coordinate evidence, owners, documentation, incidents and team compliance work in one governed workspace.
START COMPLIANCE WORKSPACE →Operate broader high-risk, GPAI, FRIA, post-market and material-change governance across a growing portfolio.
START GOVERNANCE PRO →Run institutional governance with expanded users, authority workflows, examiner rooms and portfolio reporting.
START INSTITUTION →Use the free classifier to establish the system, intended purpose, possible actor role, EU exposure and unresolved facts. When continuing evidence infrastructure is needed, paid access begins at $19 per month. Independent human readiness review remains a separate service.
It can be. Annex III lists AI used to monitor and detect prohibited behaviour of students during tests as a high-risk education use. The exact intended purpose and degree of influence still matter.
No. The classification depends on what the system does. Simple video transport or administrative scheduling differs from AI that detects, scores or flags student behaviour for possible misconduct.
Article 5 prohibits emotion inference in education institutions except where the system is intended for medical or safety reasons. That prohibition applies separately from the Annex III high-risk route.
A high-risk system must support effective human oversight. The responsible human should be able to examine the evidence, understand relevant limitations, avoid over-reliance on the model, and disregard or reverse the output where appropriate.
Current European Commission guidance states that rules for Annex III education systems apply from 2 December 2027.
Preserve system identity and version, intended purpose, monitored signals, detection rules or scores, relevant captured evidence, human review, overrides, notices, logs, performance and false-positive evidence, appeals or challenges, material changes and revalidation history.
No. TA-14 can preserve the classification basis, prohibited-practice analysis, review evidence, human-oversight state, gaps, appeal history and revalidation record. It does not provide legal advice, certification, conformity assessment or regulatory approval.
TA-14 Exchange Activity
Live cumulative activity recorded across the public Exchange surface.
···
Visitors
Recorded public visitors
···
Page Views
Recorded Exchange views