Before sharing interview material, map what goes into the tool, who receives it, what remains afterward and how you can request deletion. Check the resume, live audio, transcript and generated answers separately. Our recommendation is to start with fictional material and withhold real or confidential information while important handling questions remain unanswered.
The Canadian privacy authorities' generative AI principles recommend using anonymized, synthetic or de-identified data when personal information is not required, and call for appropriate retention schedules.[23] Their guidance is addressed to organisations and is not a blanket legal ruling about every candidate's personal use.[23] The worksheet below is a practical review aid, not a certification or legal opinion.
Map one interaction from input to deletion
Choose the exact feature you intend to use. Do not try to approve “AI” as a single category. A resume upload and a live transcription session may raise different questions even when they appear in the same interface.
Write a simple sequence: input → processing → saved output → later use → deletion. For each arrow, ask who controls the next step. If a provider's documentation does not answer, write “unknown” rather than substituting an assumption such as “probably local.” A desktop window, a browser tab and a padlock icon are not complete answers to this worksheet.
InterviewRep's existing guide says it processes live audio for transcription and uses interview text plus supplied context for suggestions; saved sessions retain transcript and answer messages.[20] The same guide explicitly does not certify processor retention, training practices, deletion coverage or compliance.[20] Those limits are important: the existence of a delete button is not enough evidence to declare every downstream copy removed.
The data-flow worksheet
Use this original table for one vendor and one workflow at a time. Fill the last column from dated documentation or a specific vendor response, not from a slogan.
| Data category | Question to answer | Evidence to record |
|---|---|---|
| Resume and role context | Is the original file uploaded, extracted text stored, or both? | Document URL, policy date, named storage or processing recipient |
| Live audio | Which voices and audio sources are captured, and where is transcription performed? | Feature-specific explanation; distinguish microphone from remote audio |
| Transcript | Is it saved automatically or only by choice, and for how long? | Saving control and retention period or retention-ending event |
| Generated suggestions | Are prompts and answers retained separately from the transcript? | History, logging and processor-handling explanation |
| Operational records | Can diagnostic logs contain interview text or uploaded material? | Policy or support answer about content in logs and access |
| Deletion and backups | What is removed immediately, what remains and when does removal finish? | Scope, exceptions, backup window and request/confirmation process |
Add a separate row if the tool has a screenshot or file feature you plan to use. Do not assume that feature exists in InterviewRep: this is a vendor-neutral worksheet. Likewise, do not infer that every vendor collects every category listed here. These are questions, not findings about a specific service.
Ask about training separately from service delivery
Ask: “Is my input or output used to train or improve models? Which organisation does that, under which setting or agreement, and does an opt-out apply to the product and its processors?” Record the actual answer and where it applies.
The Canadian principles call for openness about secondary purposes, including when personal information from prompts is used for further training or refinement.[23] That supports asking a specific question; it does not establish that any particular vendor trains on your data or that an API automatically prevents all retention.
Keep three answers distinct: what is needed to provide the feature, what is retained for operation or support, and what is used for training or improvement. A promise about one should not be copied into the other two worksheet cells.
Reduce the input before testing
Our recommendation is to create a working copy of your resume with only the facts needed for the task. Remove a home address, private contact details and other material that is irrelevant to practising an answer. Preserve the real work facts you need to discuss; do not replace them with invented achievements.
For a first test, use a wholly fictional resume and scripted conversation. Do not use a real past interview transcript merely because it is already on your computer. It may contain another person's words or details about a company that do not belong in a vendor trial.
For confidential projects, practise with a description of the problem and your contribution that you are authorised to share. If you cannot tell whether a detail is restricted, leave it out until you can check. Changing a customer name alone is not our recommended basis for deciding that a document is safe to submit.
A fictional example: one vague answer is not enough
Suppose a fictional vendor says, “Audio is deleted after transcription.” A candidate marks the entire worksheet “safe.” The missing questions are the transcript, resume, generated output, logs, downstream processing and backups. Audio deletion addresses only the audio row.
A better follow-up is: “Thank you. How long are the resulting transcript and generated answers retained? Which processors receive them, and what does deleting a session remove from their systems and backups?” Save the reply with the reviewed policy date. If the response still leaves a material gap, keep the row unknown.
This example is not an allegation about InterviewRep or another named provider. It demonstrates how to evaluate the scope of an answer without turning a partial statement into a broad privacy claim.
Test visible controls without mistaking them for an audit
Run a small synthetic test, then record only what you actually observe. Can you find the saved item? Can you export it if that is a documented feature? Does it disappear from the visible history after you use the deletion control? Can you locate the support route for a handling question?
Those checks can establish visible interface behaviour. They cannot establish that processor copies, backups or logs have been deleted. Our recommendation is to label your evidence accordingly: “no longer visible in history” is narrower and more honest than “fully erased everywhere.”
Do not repeatedly start a metered session just to explore controls without checking usage terms. InterviewRep's public page says each started minute of a live transcription connection consumes a credit, and resuming starts a new charged minute.[16] This is a usage consideration, not a privacy finding.
Decide what the unanswered questions mean
Use this practical stop/go note after completing the worksheet:
- Proceed with a limited synthetic test: the goal is learning the workflow, no real interview data is needed, and the test itself is within the rules.
- Withhold real material: recipients, retention, training or deletion remain unclear for information you would not be comfortable sharing under those unknowns.
- Do not use the workflow: the interview rules prohibit it, you lack the required permission, or the material is not yours to submit.
A completed worksheet is not a guarantee of security. Recheck it when a feature, policy, processor or intended use changes. Our recommendation is to preserve the date and the specific workflow so a later review does not inherit an approval meant for a different feature.
Common questions
Does permission to use AI also permit capturing the interviewer?
Ask about capture separately. Our permission request templates make that distinction explicit. This article does not determine the consent requirements or recording laws applicable to your situation.
Is removing my name enough?
Do not rely on that alone. Review the whole input for information you are not authorised or willing to share. For an initial trial, a fully fictional example avoids needing to decide whether a partially edited real document remains identifying.
Can InterviewRep certify this checklist for me?
No certification is offered here. The existing live-assistance guide describes public workflow facts and limits; use current policies and specific vendor answers for unresolved data-handling questions. The rule decision card addresses whether live use is appropriate in the first place.
Evaluate before uploading
If the rules and your data review permit the intended workflow, explore InterviewRep with non-sensitive context first. Its public page offers new accounts 10 free minutes with no credit card required.[16] A free offer does not reduce the need to understand what you are sharing.
Method and limits
By InterviewRep, the product publisher. Reviewed November 1, 2026. We consulted the public Canadian privacy principles and InterviewRep's current homepage and live guide. We did not inspect backend systems, perform penetration testing, verify processor deletion or obtain a legal assessment. The worksheet, fictional example and recommended stop/go process are original editorial tools, not evidence that a vendor has passed a security audit.
Sources
- InterviewRep | Live interview help, in your own words
- Live interview assistance: how it works and what to check | InterviewRep
- Principles for responsible, trustworthy and privacy-protective generative AI technologies - Office of the Privacy Commissioner of Canada
10 minutes for new accounts. No credit card required. Use only where assistance and capture are permitted. Check current plans and terms.