What the laws actually require of AI resume screeners and video interviews — NYC's bias-audit law, Illinois, Colorado, the EU AI Act, and EEOC enforcement — and how thin compliance still is.
Automated tools now sit between most applicants and a recruiter: resume parsers rank candidates, chatbots screen them, and video-interview software scores them. A wave of regulation has tried to put guardrails on this — bias audits, disclosure notices, impact assessments. But there is a large gap between what the statutes say and what is actually enforced. This page collects what the binding rules require, what the first enforcement actions looked like, and what independent studies found when they checked whether anyone was complying.
Effective July 2023 and enforced by NYC's Department of Consumer and Worker Protection, the law bars use of an "automated employment decision tool" unless it has had an independent bias audit within the prior year, the audit summary is posted publicly, and candidates get advance notice. Civil penalties run $500 to $1,500 per violation per day.
A FAccT 2024 study ("Null Compliance") examined 391 NYC employers and found only 18 had posted a required bias-audit report and only 13 had posted the candidate transparency notice. The authors concluded the law's design lets employers decide whether it even applies to them, making compliance effectively optional.
Sources: [3]
Reviewing the same 32 companies DCWP had checked, state auditors found 17 instances of potential non-compliance where DCWP had flagged only one. The audit also found 75% of test calls to NYC's 311 line about these tools were misrouted and never reached the agency, and that DCWP received just two complaints over the two-year review period.
Sources: [4]
The Artificial Intelligence Video Interview Act (effective 2020) makes employers notify applicants that AI may analyze their video, explain in general terms how it works, and obtain consent first. Employers who rely solely on AI to decide who advances must report applicant race and ethnicity data to the state, which checks the data for racial bias.
Sources: [5]
Annex III designates AI used to advertise jobs, filter applications, and evaluate candidates as high-risk, triggering risk management, data-governance, human-oversight, logging, and transparency duties. Violations of high-risk obligations carry fines up to €15 million or 3% of global annual turnover, whichever is higher.
Sources: [6]
In 2023 iTutorGroup paid $365,000 to settle EEOC claims that its software auto-rejected female applicants 55+ and male applicants 60+, screening out more than 200 people. Separate 2023 EEOC technical assistance confirmed employers are liable for vendor tools and that the four-fifths rule (a selection rate below 80% of the top group's) is a starting screen for adverse impact.
We relied on the statutes and official enforcement materials themselves (NYC DCWP, the EEOC, the Illinois Compiled Statutes, the EU AI Act text) plus peer-reviewed compliance research and a state-government audit. Where a law's status was in flux — notably Colorado's AI Act, whose effective date has been repeatedly postponed and which was substantially revised by SB 26-189 in 2026 — we describe the regulatory direction rather than a fixed requirement.
For job seekers the practical takeaway is sober: regulation exists, but the NYC evidence shows enforcement is weak and disclosure is rare, so you usually will not know whether a tool screened you or whether it was audited. The rules that bite hardest are the old-fashioned ones — the iTutorGroup case turned on a blunt age cutoff that any anti-discrimination statute would catch. The defensible response is the same as it has always been: present clean, verifiable qualifications, and where a jurisdiction grants you a notice or data-access right, use it.
How many resumes really get rejected before a human sees them, who's running the systems, and what the published data does — and doesn't — say.
Bertrand-Mullainathan, the 2023 NBER replication, age discrimination audits, and what the field-experiment literature actually shows about hiring bias.
Survey data on how candidates use AI to write resumes, how often employers can tell, and what — if anything — they do about it.
Who negotiates, who gets penalized for it, and how much it explains the pay gap — from the classic "women don't ask" data to field experiments and newer studies showing the picture has shifted.
The data shows what works. Build the resume that benefits from it.