AI Interviewers in Hiring: 12 Compliance Questions HR Should Answer Before Going Live

Quick answer: What should HR check before using an AI interviewer in 2026? Before an AI system conducts, records, scores or evaluates candidate interviews, HR should know exactly what the system does, what candidate data it uses, whether it influences who advances, what candidates are told, whether consent is required, how human review works, what testing and logs exist, and whether any features analyse emotions, facial expressions, voice or other sensitive signals.

AI interviewers are moving from experimental recruiting technology into mainstream hiring workflows.

Some systems now conduct first-round interviews through chat, voice or video. Others ask adaptive follow-up questions, generate transcripts, score answers, produce candidate summaries, recommend who should progress or write results directly back into an applicant tracking system.

That can save recruiter time. It can also change the legal and governance profile of what used to be a straightforward interview.

The important question is no longer simply whether your company uses an AI interviewer.

The important question is what the AI interviewer is allowed to observe, infer, score and influence.

Why AI interviewer compliance is becoming a real HR issue

AI interview technology is becoming more capable at the same time that employment-related AI rules are becoming more specific.

A basic scheduling assistant presents one level of risk. A conversational system that asks candidates questions is different. A system that also evaluates answers, assigns scores, infers traits, ranks applicants or automatically advances candidates moves closer to consequential decision support.

This is why HR teams should evaluate AI interviewers as a workflow, not as a single feature.

For a broader overview of the regulatory landscape, see HRYP’s AI Hiring Compliance in 2026 guide. If you are still selecting a product, use the AI Recruiting Vendor Due Diligence Checklist alongside this interview-specific review.

What is an AI interviewer?

An AI interviewer is a system that interacts with candidates during part of the interview or screening process using artificial intelligence.

Depending on the product, it may:

• conduct text, voice or video interviews;
• ask structured or adaptive follow-up questions;
• record or transcribe candidate responses;
• summarise interviews for recruiters;
• score answers against a rubric;
• identify skills or job-related signals;
• rank candidates or recommend who should progress;
• trigger an automated next step in the ATS.

These functions should not automatically be treated as equivalent. A transcription tool and an autonomous screening agent may sit inside the same recruiting stack but create very different governance questions.

Is it legal to use AI for job interviews in 2026?

AI interviewing is not universally prohibited. But the answer depends on the jurisdiction, the system’s functionality and how its output is used.

Some laws focus on candidate notice or consent. Others regulate automated employment decision tools, discrimination, high-risk AI, biometric processing, transparency or human oversight.

For example, New York City’s Local Law 144 can apply when an automated employment decision tool substantially assists or replaces discretionary decision-making in hiring or promotion. Illinois has a specific Artificial Intelligence Video Interview Act for certain AI-analysed video interviews, as well as separate 2026 employment AI provisions under the Illinois Human Rights Act. In the European Union, certain AI systems used to evaluate candidates can fall within the AI Act’s high-risk framework, while separate prohibited-practice and transparency rules already apply.

That is why an employer should identify the exact interview functionality before trying to answer the legal question.

AI Interviewer Compliance Checklist: 12 Questions HR Should Answer

1. What exactly does the AI interviewer do?

Start with a plain-language description of the workflow.

Does the tool only ask predetermined questions? Does it choose follow-up questions dynamically? Does it record video or audio? Does it analyse the content of answers? Does it generate a score? Does it recommend a hiring action?

Do not accept a label such as “AI-powered interview” as a sufficient description.

HR should be able to draw the workflow from candidate invitation to recruiter decision.

2. Does the system only conduct the interview, or does it evaluate the candidate?

This distinction is critical.

A system may simply facilitate an interview and hand the transcript to a recruiter. Another system may interpret the answers and generate a candidate score. A third may use that score to determine who moves forward.

Ask whether the AI output:

• changes candidate priority;
• creates a pass/fail result;
• ranks candidates;
• recommends rejection or advancement;
• automatically triggers the next recruiting step.

The closer the system gets to influencing a consequential employment decision, the more carefully the organisation should assess applicable law, testing, documentation and human oversight.

3. Are candidates clearly told they are interacting with AI?

Candidate transparency should be designed into the workflow rather than added as an afterthought.

In the EU, Article 50 of the AI Act includes transparency obligations for providers of AI systems intended to interact directly with natural persons. Those transparency provisions became applicable on 2 August 2026. The exact allocation of responsibility depends on the system and the organisation’s role, but an HR team should verify that candidates receive clear information rather than assuming the vendor has handled the issue.

A practical disclosure should answer basic questions such as:

• Is this interview being conducted by AI?
• Will the interview be recorded?
• Will AI analyse the candidate’s responses?
• Will the AI produce a score, recommendation or ranking?
• Will a human review the result?

4. Is candidate consent required, and can you prove it was obtained?

Notice and consent are not the same thing.

One of the clearest examples is Illinois. The Illinois Artificial Intelligence Video Interview Act requires an employer that asks applicants to record video interviews and uses AI to analyse those videos for positions based in Illinois to notify applicants before the interview, explain how the AI works and what general types of characteristics it evaluates, and obtain consent before using AI to evaluate the applicant.

Employers should therefore know which candidates and roles trigger consent requirements and should preserve evidence that the required consent step actually occurred.

5. What is recorded, transcribed, stored or sent to third parties?

An AI interview can create much more data than a conventional phone screen.

Potential data may include:

• audio recordings;
• video recordings;
• transcripts;
• generated summaries;
• question-and-answer history;
• candidate scores;
• metadata about timing or interaction;
• derived attributes or classifications.

Ask the vendor what is stored, where it is stored, who can access it, how long it is retained, whether subprocessors receive it and whether customer or candidate data is used to train or improve models.

6. Does the system infer emotion, enthusiasm, personality or similar signals?

This is one of the most important checks for employers operating in or hiring into the European Union.

Article 5 of the EU AI Act prohibits the use of AI systems to infer emotions of natural persons in the workplace, except for limited medical or safety reasons. The European Commission’s guidance interprets the workplace concept broadly and expressly states that it includes candidates during the selection and hiring process. The guidance gives the example that using emotion-recognition AI during recruitment is prohibited.

The prohibited-practice rules have applied since 2 February 2025.

That means HR should not rely on marketing language such as “emotion aware”, “sentiment scoring”, “enthusiasm detection” or similar claims without understanding exactly what the system infers and whether that functionality is lawful for the intended use.

Do not confuse analysing the semantic content of an answer with inferring a person’s emotional state from biometric or behavioural signals. They can raise different legal questions.

7. Which inputs influence the candidate score or recommendation?

A score is only as understandable as the factors behind it.

Ask whether the system evaluates:

• the factual content of answers;
• job-related competencies;
• keywords or semantic similarity;
• speech characteristics;
• facial or visual information;
• response timing;
• historical hiring data;
• personality or culture-fit proxies;
• other inferred attributes.

If neither HR nor the vendor can explain which inputs materially affect the output, it becomes difficult to validate, challenge or govern the system.

8. Can a human meaningfully review and override the result?

“Human in the loop” should describe an actual control, not a slogan.

Ask whether a recruiter can:

• review the underlying interview or transcript;
• see how the AI reached its recommendation;
• override a score or recommendation;
• restore a candidate screened out by automation;
• stop an automated workflow before rejection;
• document why the human disagreed with the AI.

A nominal override button is not enough if the system hides the evidence a recruiter would need to exercise independent judgment.

9. What testing exists for bias, accuracy and job relevance?

Ask for evidence, not adjectives.

Useful vendor questions include:

• What exactly was tested?
• Which product version was tested?
• Which roles or populations were included?
• Was the testing internal or independent?
• What performance limitations were identified?
• How often is testing repeated?
• What happens after a material model or scoring change?

For employers covered by New York City’s Local Law 144, an AEDT cannot be used unless it has undergone the required bias audit within one year of use, the required audit information is publicly available and the required notices are provided.

10. What logs and evidence will exist after an interview?

If a candidate challenges an outcome six months later, can the organisation reconstruct what happened?

Useful records may include:

• the system and model version;
• the interview questions asked;
• the candidate’s responses;
• the score or recommendation produced;
• the criteria or rubric in use;
• the recruiter who reviewed the output;
• any human override;
• the final decision and workflow step.

Retention decisions should also be reviewed against privacy, employment, litigation-hold and jurisdiction-specific requirements rather than keeping every AI-generated artifact indefinitely.

11. What happens when the vendor changes the model?

An AI interviewer can change during the life of a contract.

The vendor may update the language model, scoring method, interview logic, anti-cheating features, integrations or underlying data-processing architecture.

HR should know:

• how material changes are communicated;
• whether new features are enabled automatically;
• whether scoring changes trigger new validation or testing;
• whether customers can delay or disable an update;
• whether the organisation can identify which version affected a particular candidate.

This becomes even more important as recruiting products evolve from AI assistants into more autonomous agents. HRYP’s guide to AI Agents in HR explains why governance needs to cover actions and permissions, not only prompts.

12. Can the organisation pause the AI without breaking recruiting?

Every consequential AI workflow needs an off-ramp.

If legal guidance changes, a bias issue appears, the vendor releases an unexpected model update or the system begins producing unreliable results, can HR temporarily disable AI interviewing and continue hiring?

A resilient process should have a manual fallback and a clear owner with authority to suspend the AI feature.

AI interview compliance by jurisdiction: three areas HR should know

European Union: AI Act rules for interviews and candidate evaluation

The EU AI Act treats certain AI systems used in employment and recruitment as high-risk when they are intended for use cases such as analysing or filtering job applications or evaluating candidates.

Following the AI Omnibus changes that entered into force in July 2026, the rules for Annex III high-risk systems are scheduled to apply from 2 December 2027.

That later date should not be confused with other AI Act provisions that are already applicable.

Two current examples matter for AI interview workflows:

• Prohibited practices: the prohibition on emotion recognition in the workplace has applied since 2 February 2025, and Commission guidance expressly includes recruitment candidates within that interpretation.
• Transparency: Article 50 transparency obligations for certain systems, including AI intended to interact directly with natural persons, became applicable from 2 August 2026.

Not every tool used during an interview is automatically a high-risk system. The intended purpose and actual functionality matter. A scheduling chatbot, transcription assistant and candidate-evaluation engine should therefore not be collapsed into one category.

Illinois: AI video interviews and wider 2026 employment AI rules

Illinois has two separate areas HR teams should distinguish.

The state’s Artificial Intelligence Video Interview Act has been in force since 2020 and applies to certain employer-requested recorded video interviews that are analysed by AI for positions based in Illinois. It requires advance notice, information about how the AI works and the general characteristics it evaluates, and applicant consent before AI evaluation.

Separately, Public Act 103-0804 took effect on 1 January 2026 and amended the Illinois Human Rights Act to address the use of AI in recruitment and other employment decisions. It prohibits AI use that has the effect of subjecting employees to discrimination based on protected classes and establishes a notice requirement, with implementation details subject to Illinois Department of Human Rights rulemaking.

Because rulemaking can evolve, employers should verify the current IDHR rules and guidance before relying on a specific notice format or process.

New York City: when an AI interviewer becomes an AEDT issue

New York City’s Local Law 144 does not apply to every piece of recruiting software merely because it contains AI.

It applies to automated employment decision tools that fall within the law’s definition and are used in covered employment decisions.

Where the law applies, the employer or employment agency cannot use the AEDT unless a bias audit has been conducted within one year, required information about that audit is publicly available, and required notices have been provided.

An AI interviewer that merely records a conversation may be different from one that scores the interview and substantially assists who progresses. The workflow determines the issue.

A practical AI interviewer go-live process

HR does not need to solve every AI governance problem before evaluating a product. But it should complete a minimum operational review before allowing the system to influence live candidates.

1. Inventory the feature. Document whether the system conducts, records, transcribes, scores, ranks, recommends or advances candidates.
2. Map the decision path. Identify what happens immediately before and after the AI output.
3. Identify jurisdictions. Determine where roles and candidates are located and which rules may apply.
4. Review candidate notice and consent. Make sure required disclosures occur before the relevant interaction.
5. Check prohibited or sensitive features. Pay particular attention to emotion recognition, biometric analysis and opaque personality inference.
6. Validate human oversight. Confirm a recruiter can independently review and override consequential outputs.
7. Request evidence from the vendor. Obtain testing, documentation, data-use and change-management information.
8. Define logs and retention. Decide what must be kept and who can access it.
9. Pilot before scaling. Test representative roles and review unexpected outcomes before wider deployment.
10. Create a pause procedure. Know who can disable the AI and how recruiting continues without it.

AI interviewer red flags HR should not ignore

Some warning signs deserve escalation before launch:

• The vendor cannot clearly explain whether the AI scores or ranks candidates.
• The system evaluates “emotion”, “enthusiasm”, “confidence” or similar signals without a precise technical and legal explanation.
• Candidate disclosure is vague or appears only inside a long privacy policy.
• The organisation cannot determine when consent is required.
• Recruiters can see a score but not the evidence behind it.
• Candidates can be automatically rejected without meaningful human review.
• The vendor provides no useful testing or version information.
• Model updates can change scoring without customer review.
• No one inside the organisation owns the decision to suspend the system.

These red flags do not automatically mean the product is unlawful or unusable. They mean the organisation does not yet have enough information to treat the deployment as controlled.

AI interviewer vs AI interview assistant: why the distinction matters

HR teams should avoid treating every AI interview tool as if it performs the same role.

This functional approach also reduces procurement confusion. Instead of asking whether a product is “AI compliant”, HR can evaluate the specific features that touch candidates and decisions.

Frequently asked questions about AI interviewers and compliance

Do candidates have to be told that an interview is being conducted by AI?

Often they should be, and in some situations a legal requirement applies. In the EU, Article 50 creates transparency obligations for providers of AI systems intended to interact directly with people, subject to its scope and exceptions. Other jurisdictions, including Illinois and New York City, impose notice requirements in specific covered AI hiring scenarios. The exact obligation depends on the system, the organisation’s role and the jurisdiction.

Do candidates have to consent to an AI interview?

Not under one universal global rule. Consent requirements depend on the jurisdiction and type of system. Illinois provides a clear example: employers covered by the Artificial Intelligence Video Interview Act must obtain consent before using AI to evaluate covered applicant-submitted video interviews.

Can AI analyse facial expressions or emotions during a job interview?

This requires particular caution. Under the EU AI Act, AI systems used to infer emotions in the workplace are prohibited except for limited medical or safety reasons. European Commission guidance states that the workplace concept includes candidates during the selection and hiring process and gives AI emotion recognition during recruitment as a prohibited example.

Does the EU AI Act ban AI interviews?

No. The AI Act does not impose a blanket ban on AI interviewing. Certain prohibited practices are banned, while other recruitment and candidate-evaluation systems may fall within the high-risk framework depending on their intended purpose and functionality.

When do EU high-risk AI hiring rules apply?

Under the current EU implementation timeline, the rules for high-risk AI systems listed in Annex III are scheduled to apply from 2 December 2027. Other AI Act provisions already apply on different dates, including prohibited-practice rules and Article 50 transparency obligations.

Does NYC Local Law 144 apply to every AI interviewer?

No. The question is whether the tool falls within the law’s definition of an automated employment decision tool and is used in a covered way. A system that scores or recommends candidates may raise different issues from a tool that only records an interview for later human review.

Is human review enough to make an AI interviewer compliant?

No single control guarantees compliance. Human review can be important, but organisations may also need to address notice, consent, testing, bias audits, data processing, documentation, prohibited features or other requirements depending on the system and jurisdiction.

What should HR ask an AI interviewer vendor before purchase?

Ask what the system evaluates, whether it scores or ranks candidates, what data it records, what testing exists, how humans override outputs, what logs are available, how model changes are managed and whether sensitive features such as emotion or biometric inference are used. HRYP’s 15-question vendor due-diligence checklist provides a broader procurement framework.

What is the fastest way to review an existing AI interview process?

Start by mapping the interview workflow from candidate invitation to hiring decision. Identify each AI system, what data it sees, what output it produces, whether that output changes candidate progression and where meaningful human review occurs. This is the same process-oriented approach used by the HRYP AI Hiring Readiness Check.

The bottom line

AI interviewers can remove a significant amount of repetitive screening work, especially in high-volume recruiting.

But the biggest risk is not simply that the interview is automated.

The bigger risk is allowing a system to record, infer, score or influence candidate outcomes without the organisation clearly understanding what is happening.

Before an AI interviewer goes live, HR should be able to answer five basic questions:

1. What does the system actually do?
2. What does the candidate know?
3. What data and signals does the AI use?
4. How does a human review or override the output?
5. What evidence can the organisation produce later?

If those answers are unclear, the organisation is not ready to treat the AI interviewer as an ordinary recruiting tool.

Official sources and regulatory references

• European Commission — AI Act regulatory framework and current implementation timeline
• European Commission — AI Omnibus enters into force and updated high-risk timelines
• European Commission AI Act Service Desk — Article 5 prohibited AI practices
• European Commission AI Act Service Desk — Article 50 transparency obligations
• New York City Department of Consumer and Worker Protection — Automated Employment Decision Tools / Local Law 144
• Illinois General Assembly — Artificial Intelligence Video Interview Act, Section 5
• Illinois General Assembly — Public Act 103-0804, AI use in employment

Last updated: September 2026. This article provides general HR technology and compliance information and does not constitute legal advice. AI, employment, privacy and discrimination rules vary by jurisdiction and continue to evolve. Organisations should obtain appropriate professional advice where a legal determination is required.