Is It Legal to Detect Cheating in Video Interviews?
Short answer: yes, with rules. Longer answer: the rules are getting stricter, and they differ by where you hire.
This is practical guidance, not legal advice. Talk to your counsel before you deploy anything. What follows is what your counsel will want to see already handled.
The core principle: detect, disclose, decide
Regulators are converging on one idea. You can use AI in hiring, including fraud detection, if you tell candidates what you are doing, check the system for bias, and keep a human making the final call.
Detection without disclosure is where companies get hurt. A candidate who learns after the fact that their screen was captured, their voice was biometrically analyzed, and an algorithm scored their honesty will not care how good your intentions were.
New York City: Local Law 144
NYC's law covers automated employment decision tools used in hiring. If your fraud detection feeds a score that influences hiring decisions for NYC candidates, you likely need an independent bias audit and you must notify candidates at least 10 business days before use.
The audit requirement is the sharp edge. It applies to tools that "substantially assist or replace" human decision-making. A fraud risk score that auto-quarantines candidates is in scope. A flag that a recruiter reviews before acting is easier to defend, which is another reason to keep humans in the loop.
European Union: the AI Act
The EU AI Act classifies AI systems used in employment, including recruitment and candidate evaluation, as high-risk. High-risk systems face transparency obligations, data governance requirements, human oversight mandates, and conformity assessments.
Fraud detection in interviews sits squarely in this frame. If you hire in the EU, your vendor should be able to show how the system meets these obligations: what data it processes, how the model was validated, and where the human oversight sits. Ask for this documentation before you sign, not after a regulator does.
Candidate consent and disclosure
Everywhere else, the baseline is consent and disclosure. Tell candidates before the interview what will be captured: screen, camera, audio, and what the analysis looks for. Get explicit consent. Make it plain language, not a 40-page policy.
Three disclosure rules that keep you safe:
- Disclose before, not after. Consent at the interview invite, not in the rejection email.
- Be specific. "We analyze your interview for signs of AI-assisted cheating, including screen activity, voice patterns, and video authenticity." Vague notices invite disputes.
- Offer an alternative path. A candidate who declines screen capture should still be able to interview another way. The alternative can be slower or more manual. It must exist.
Recordkeeping is your shield
Every flag needs a record: what was observed, when, at what confidence, and who made the final decision. Regulators and courts do not punish detection. They punish detection you cannot explain.
Keep the evidence trail per candidate. Keep model versions logged, so a flag from March can be reconstructed in November. Keep the recruiter's decision attached to the flag. A complete record turns a dispute into a demonstration.
RightMatch logs every signal, every score change, and every action with timestamps, per candidate. The audit trail is built for EEOC reviews, bias audits, and compliance reporting.
The checklist
Before you deploy interview fraud detection:
- Disclose capture and analysis to candidates in advance, in plain language
- Get explicit consent, with an alternative interview path available
- Confirm bias audit coverage where required (NYC and similar jurisdictions)
- Confirm your vendor documents EU AI Act high-risk obligations if you hire in the EU
- Verify the evidence trail: timestamped, explainable, auditable per flag
- Keep the final decision with a human, always
The line: Regulators do not punish detecting cheating. They punish detecting it without disclosure, without a bias audit, and without a record. Disclose, audit, and log everything.