Policy
What You're Agreeing To in a Candidate AI-Use Policy
Five things usually sit inside a candidate AI-use agreement: what counts as assistance, which stages the rule covers, whether your screen or session is recorded and for how long, an attestation that your submission is your own, and what happens if the rule is broken, sometimes reaching past the interview into an offer already made. Most of it is standard and worth signing. Ask about anything unusual before you do, especially recording and retention, since a clarifying question here is normal and isn't read as resistance.
The takeThese documents get drafted once, by legal or HR, and then get read cold by hundreds of different candidates who have no baseline for telling a standard clause from an unusual one. A retention period that runs past the hiring decision, a recording scope that includes the desktop rather than one tab, a broken-rule consequence that reaches back after an offer: these vary more than candidates assume, and the only way to catch the unusual one is to read the ordinary version first.
Where Olive fits
Open a role and see what the work shows
An Olive invite carries its own terms in plain language rather than a separate document to sign: what's captured is the assessment tab only, never a camera, and the six findings a person writes afterward are handed to you in the identical copy your employer receives.
Rank your shortlistWhat am I actually agreeing to when I sign this?
You're agreeing to a set of terms an employer wrote for every candidate in this process. Most of these documents cover the same five areas: what counts as AI assistance, which stages the rule governs, what gets recorded and how long it's kept, a statement that your submission is genuinely yours, and what happens if the rule is broken. What varies is the specifics inside each clause, and that's where reading matters.
Treat it like any other document with your name on it: read every clause once before you sign, note anything that surprises you, and ask about that specific line rather than the whole document. A candidate who asks one focused question about retention or scope reads as careful, not difficult, and a well-run process expects a few of these questions every cycle.
If nothing in the document surprises you and the scope looks like what you'd expect from an ordinary hiring process, sign it and move on. The point of reading closely isn't suspicion. It's knowing exactly what you agreed to if the topic comes up again later in the process.
Read the five clauses closely before you sign
Start with the definition of assistance, since it sets the scope for everything that follows. A document that defines assistance narrowly, generating text or code with a named tool, is easier to comply with than one that reaches into research, scheduling or translation without saying so. If the definition reads vague, ask for one example of something that counts and one that doesn't.
| Clause | What it usually says | What's worth a question |
|---|---|---|
| Assistance | Names which tools or acts count | A definition vague enough to cover anything |
| Stages covered | Which rounds the rule applies to | A rule that doesn't say, or contradicts an earlier email |
| Recording and retention | What's captured and how long it's kept | Desktop or camera capture, or no stated deletion date |
| Attestation | A statement that your submission is your own | Wording broad enough to cover normal editing help |
| Consequence | What happens if the rule is broken | Language reaching past the interview into an offer |
The attestation clause is usually the shortest line in the document and the easiest to skim past, which is a mistake, since it's the one that turns an inaccurate answer into a real problem later rather than an ordinary one. Read exactly what you're confirming: that your submission reflects your own reasoning and judgment is a fair thing to sign. Wording broad enough to disclaim any AI contact at all is worth a direct question, since most employers don't actually mean that.
Recording and retention is the clause most worth reading twice, because it's the one part of this document that state law already regulates in places. Illinois requires an employer to tell you before an AI-analyzed video interview, explain roughly how the analysis works, and get your consent before evaluating you with it 1, and if you ask, the employer must delete your video within 30 days and instruct anyone who received a copy to do the same 2. What an employer is required to disclose about screen and session recording more broadly covers the same ground from the other chair, and it's worth reading if the agreement in front of you is vague about retention.
Which state laws already apply to this document?
A few US jurisdictions already regulate parts of what these agreements describe. Illinois has required consent before AI analysis of a recorded video interview since 2020, and gives you a right to withhold that consent, though the statute sets no penalty for ignoring a refusal 3. Maryland has required a signed waiver since the same year before a facial recognition service can create a facial template during your interview, naming you, the date, your consent, and whether you read it 4.
New York City takes a different approach for automated tools that substantially help decide who advances: an employer must have run a bias audit within the past year, posted a public summary of it, and given you at least ten business days' notice before the tool is used, in force since January 2023 5. None of these laws requires the agreement in front of you to look any particular way. They do mean a clause promising notice, consent or a retention limit is often restating something the law already requires, which is a reasonable thing for a document to do rather than a red flag on its own.
Don't treat a clarifying question as resistance
Employers running a real process expect a few candidates to ask about the document before they sign it, and a specific question about one clause reads as diligence rather than suspicion. Ask what happens to the recording if you're not hired, whether the definition of assistance includes research or only generation, and what a broken rule would actually look like in practice for this role.
The consequence clause deserves the most direct question of the five, because what happens when a violation surfaces after an offer is already out is a real and researched decision on the employer's side, not a hypothetical. If the document you're signing reaches that far, asking how a violation would actually be identified and handled is a fair thing to raise before you sign, not after something has already gone wrong.
This is ordinary due diligence. Most candidate AI-use agreements exist because an employer wants a clear, repeatable answer to a question it used to handle inconsistently in email threads and verbal asides, and a written policy applied the same way to every candidate is usually the fairer version of that same rule. Reading it carefully and asking one good question is how you confirm it's actually being applied that way in your case.
Common questions
Is it normal to be asked to sign something like this before an interview?
Increasingly, yes. As more employers formalize rules about candidate AI use, a written agreement is becoming a standard part of the process rather than a sign that something unusual is happening to you specifically. Read it the way you'd read any onboarding paperwork: once, carefully, with questions about anything unclear.
What if I don't want to sign it?
That varies by employer, and the document itself usually says what happens if you decline, sometimes nothing, sometimes that you can't proceed to that stage. If it doesn't say, ask directly rather than assuming either outcome. Declining to sign is a real option worth understanding the consequence of before you use it.
Does signing this mean my whole screen or camera is recorded?
Not necessarily, and it's worth confirming rather than assuming. Well-built agreements name exactly what's captured, often one browser tab rather than the desktop, and state plainly whether a camera is involved at all. If the document is silent on scope, that silence is the specific thing to ask about before you sign.
What if the document doesn't say how long recordings are kept?
Ask for a number. A well-built policy states a retention period and a deletion process; the absence of either is common enough that asking rarely reads as unusual. Illinois law, where it applies, requires an employer to delete an AI-analyzed interview video within 30 days of the candidate's request, which is a reasonable benchmark to ask an employer to match even outside that state.
Can I keep a copy of what I signed?
Ask for one if it isn't offered automatically. A signed agreement affects you, and having your own copy of exactly what you agreed to is reasonable to request and easy for an employer to provide. If an employer won't give you a copy of something you signed, that's worth noting on its own.
References
- 1. Artificial Intelligence Video Interview Act, 820 ILCS 42 ilga.gov Illinois requires notice, an explanation of the AI, and consent before an AI-analyzed video interview.
- 2. 820 ILCS 42, Artificial Intelligence Video Interview Act, Sec. 15 (Destruction of videos) ilga.gov On request, the employer must delete an Illinois AI-interview video within 30 days, including copies held by others.
- 3. 820 ILCS 42, Artificial Intelligence Video Interview Act, Sec. 5 (Disclosure of the use of artificial intelligence analysis) ilga.gov An applicant may withhold consent to AI video analysis, though the Act sets no penalty for ignoring a refusal.
- 4. Chapter 446 (House Bill 1202), Labor and Employment - Use of Facial Recognition Services - Prohibition, Md. Code Lab. & Empl. section 3-717 mgaleg.maryland.gov Maryland requires a written, four-item consent waiver before facial recognition can be used during an interview.
- 5. Automated Employment Decision Tools: Frequently Asked Questions nyc.gov NYC Local Law 144 requires a bias audit, a public summary, and ten business days' notice before an automated tool is used, effective January 2023.
5 sources, numbered by first appearance. How Olive sources claims
General guidance, not legal advice. Hiring rules differ by state and country and change often; check anything here against your own counsel before you act on it.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.