Yes. This happens more often than most organizations realize. AI video screening tools score specific behavioral signals — speech pace, response structure, facial patterns, keyword alignment. These are measurable proxies, not direct measures of job capability. Candidates who communicate differently, non-native speakers, neurodivergent professionals, experienced executives with unconventional styles, may score poorly on metrics that have nothing to do with their actual performance. Strong candidates disappear from your funnel before a human ever sees them.

Yes. Several US jurisdictions now impose specific requirements on organizations using AI hiring tools — disclosure obligations, human review requirements, and documentation standards. Candidates have up to 300 days federally, and up to three years in Washington state, to file discrimination claims after an AI-assisted hiring decision. If your process cannot explain rejections in clear job-relevant terms or demonstrate meaningful human oversight, your legal exposure is growing as litigation and regulation accelerate.

An AI hiring audit examines what your screening tools are actually measuring versus what predicts performance in your specific roles, where your documentation gaps are, whether your human review process is meaningful or superficial, and whether your vendor's bias claims hold up against independent research. It is not a technical software review, it is an independent strategic assessment of whether your process is finding the right people and whether it can be defended if challenged.

Defensible AI hiring decisions require four things.

First: documentation of what the AI tool scores and why those signals are job-relevant.
Second: evidence of genuine human review at consequential decision points.
Third: the ability to explain any rejection in specific behavioral terms if a candidate or regulator asks.
Fourth: vendor validation that has been independently assessed not just accepted from sales materials.

Most organizations using AI hiring tools are missing at least two of these. Identifying which ones is where the work starts.

Not always, and most HR teams don't have the technical background to evaluate them independently. Vendor validation studies are designed to demonstrate their tools work, not to identify where they fail for your specific candidate population. To assess a bias claim accurately you need to understand what protected groups were included in the validation sample, what job performance outcomes the tool was validated against, and whether an independent third party reviewed the methodology, not just the vendor's own team.

Depending on the platform, AI video interview tools and AI hiring tools typically evaluate two categories of signals. Communication signals: speech pace, vocal consistency, response structure, keyword alignment, and STAR format adherence. Environmental signals: background lighting, camera quality, and facial expression patterns.

These are measurable proxies. They are not direct measures of judgment, leadership capability, or real-world performance. Understanding exactly what your specific tool scores, and what the research says about whether those signals predict job success, is the starting point of every assessment I conduct.

AI hiring regulations apply broadly across the US and internationally. What varies is which specific requirements apply to your organization based on where you operate and where your candidates are located.

New York City's Local Law 144 requires independent bias audits for automated employment decision tools. Illinois requires disclosure and consent for AI video interview analysis. Colorado's AI Act, effective January 1, 2027, imposes disclosure, human review access, and three-year recordkeeping requirements. The EU AI Act classifies hiring AI as high-risk and requires documented human oversight. Most organizations are operating across multiple jurisdictions without a clear picture of which requirements apply to them.

This landscape is evolving rapidly. What applies to your organization today may expand significantly within the next 12 months. It is important to stay on top of it to proactively protect your organization from lawsuits, regulatory complaints, and candidate disputes that waste time, drain resources, and damage your reputation as an employer.

Yes, and the research documents it clearly. AI video screening tools and AI hiring tools are typically trained on historical hiring data that reflects past patterns and often from neurotypical, native English-speaking candidates in North American professional contexts. Candidates who communicate differently due to cultural background, language, neurodivergence, or disability may score poorly on behavioral metrics that have nothing to do with their actual capability. For example, the ACLU filed EEOC charges in 2025 after an AI video screening system scored a deaf Indigenous candidate low for active listening and recommended she practice it.

Meaningful human oversight means reviewers are genuinely interpreting AI scores in context, not rubber-stamping them. It means your team understands what behavioral signals the tool scored and why those signals are or aren't relevant to the specific role. It means override decisions are documented with specific behavioral reasoning. It means someone in your organization can explain any consequential hiring decision in job-relevant terms if a candidate, regulator, or attorney asks. Most organizations have human review on paper. Fewer have it in practice. The difference matters enormously when a decision is challenged. Assessing whether your current process meets that standard is exactly what the AI Hiring Governance Review examines.

AI hiring vendors have a financial interest in selling and renewing their tools. Their validation studies are designed to demonstrate their tools work — not to identify where they fail for your specific candidate population and roles. Most HR teams don't have the technical background to evaluate psychometric validation claims, statistical bias testing methodology, or the gap between controlled study results and real-world hiring outcomes. Independent evaluation gives you the picture your vendor has no incentive to provide including whether their tool is appropriate for your specific roles, candidate pool, and organizational context.

The Human Signal Gap is the difference between what AI screening tools actually measure and what actually predicts success in your specific roles. AI tools score what they can quantify — speech pace, response structure, facial patterns, keyword density. They cannot reliably measure judgment under pressure, leadership capability, cultural context, or real-world performance. That gap between measurable signals and meaningful human capability is where your best candidates disappear before a human ever sees them and where your organization's legal and talent risk quietly accumulates.

AI-assisted interview cheating, candidates using real-time tools to feed scripted responses, is a growing problem. Detection tools can flag suspicious behavioral patterns but they produce fraud scores, not legal proof.

Acting on a fraud score without a documented protocol creates its own legal exposure. Detection tools produce false positives that can disproportionately affect non-native speakers, neurodivergent candidates, and professionals with atypical communication styles — the same groups AI screening already struggles to evaluate fairly.

The defensible approach is a documented protocol that specifies what behavioral observations trigger concern, what steps reviewers take, and critically — what interviewers can and cannot say to a candidate based on a fraud score alone. A fraud score is reasonable suspicion. It is not grounds for accusation without documented evidence.

The threat landscape extends beyond individual candidates — organized deepfake operations targeting remote technical roles have been documented at scale, making identity verification and behavioral observation skills increasingly essential components of any hiring process.

Building that protocol is something I help organizations develop as part of any audit engagement.

Every engagement starts with a free 30-minute strategy call. We discuss your specific situation, which tools you're using, and what's causing your concern. Based on that conversation I'll suggest which assessment makes the most sense or tell you honestly if something else would serve you better.