AI proctoring analyses the proctoring photos and video recordings captured during a test session and flags moments that may need attention. The analysis runs after the session on TestInvite's own servers. It never makes decisions about test-takers: every flag is a suggestion for a human reviewer.
What does the AI check for?
Each check can be turned on or off separately in the test's proctoring settings.
- No person visible. No one is in the frame, or the face is hidden.
- Multiple people. More than one person appears in the frame.
- Face position and gaze. The head is turned beyond a set angle, or the eyes look away from the screen.
- Forbidden objects. A mobile phone or a book is visible.
- Identity consistency (optional). The person in the photos is not the same person as in the identity photo taken at the start of the session, or in a reference photo uploaded by the organisation.
- Audio (video recordings only). The microphone appears muted, or voice activity is detected.
Does AI proctoring use facial recognition or biometric data?
Only the optional identity consistency check compares faces. It is a one-to-one check: the test-taker's photos are compared only with that same test-taker's identity photo or reference photo. There is no search against any database, and test-takers are never compared with each other.
For the comparison, a numeric face description is computed in memory and discarded right after. Face templates are not stored. The only things kept are the similarity score, the threshold used and the resulting flag.
The other checks (person count, head position, gaze, objects, audio) do not identify anyone. If you prefer not to use face comparison, switch off the identity consistency check and the other checks keep working.
Does the AI make decisions about test-takers?
No. Every result is a flag that your reviewers confirm or dismiss. Scores and outcomes are never changed automatically because of a proctoring flag.
Is test-taker data used to train AI models?
No. Proctoring photos, recordings, responses, results and proctoring flags are not used to train, fine-tune or improve any AI model, by TestInvite or by any third party. AI proctoring runs on pre-trained models that do not learn from the data they analyse.
Which providers process the data, and where?
No external AI service is involved in AI proctoring. Photos and recordings are analysed on TestInvite's own backend (Google Cloud Functions) and stored in Google Cloud Storage and Firestore, all in the US region. Images are not sent to any third-party AI provider.
See Data location and Technology used in TestInvite.
How long is proctoring data kept, and can it be deleted earlier?
Proctoring photos, recordings and flags belong to Supplementary Test-Taker Data. This data is deleted within one year from the date of collection.
You can delete it earlier at any time from the TestInvite application:
- Purge proctoring media. Removes the proctoring photos and recordings of a test session, or of all test sessions in an assessment, while keeping the results.
- Delete test sessions. Removes the test sessions entirely, including responses, proctoring media and flags.
Each deletion runs as a tracked job that shows its progress and how many items were removed. Because AI proctoring uses no external provider, no copies exist outside TestInvite's storage.
See Data Deletion Policies and Backups.