01
Recruiting and advertising
Targeting and sourcing determine who sees an opportunity and who enters the candidate pool.
INSTITUTIONAL RISK BRIEF · 04
AI in Labor & Employmenthiring, surveillance, and the right to challenge.
Employers use AI to find candidates, rank applicants, monitor workers, set schedules, evaluate performance, and recommend termination. Efficiency does not change the legal or institutional fact: the employer still chooses the procedure and owns its effect on access to work.
PREPARED BY SYNTHETIC OUTLAW RESEARCHSCOPE SELECTED U.S. EMPLOYMENT AUTHORITIESPUBLISHED JUL 21, 2026VERSION 1.0
01 · THE BURDEN
A vendor supplies the score. The employer still makes the employment decision.
EEOC guidance states that selection procedures can violate federal employment law when they disproportionately exclude protected groups without the required justification. Vendor documentation can help, but the employer remains responsible for ensuring that the procedure is valid for its use.
The organizational promise
AI can reduce administrative work, widen candidate discovery, improve scheduling, identify training needs, and help managers see patterns that are difficult to detect manually.
The institutional burden
The same system can turn historical preferences into a scoring rule, screen out a disability, convert behavior into a productivity proxy, or make a consequential recommendation that nobody can explain. Procurement does not transfer accountability.
02 · CONSEQUENTIAL WORKFLOWS
Where AI becomes institutional action.
The relevant question is not whether AI appears in the workflow. It is whether its output changes attention, access, price, timing, treatment, judgment, or a person’s practical ability to obtain review.
02
Screening and ranking
Resume, video, game, and assessment systems can exclude applicants before a person reviews them.
03
Performance management
Productivity scores can influence pay, promotion, discipline, and termination.
04
Scheduling and allocation
Automated schedules and task assignments can alter income, predictability, and access to preferred work.
05
Workplace surveillance
Location, communications, keystrokes, images, voice, and biometric signals can become management evidence.
03 · AUTHORITIES
The duties converge. The operating standard remains distributed.
Existing employment law applies to AI-assisted decisions. Newer local rules add audit and notice duties, but no single regime covers every tool, worker, decision, and jurisdiction.
SCOPE
Federal anti-discrimination law, EEOC technical assistance, and New York City Local Law 144 have different coverage and legal effect. A bias audit required by one jurisdiction does not establish validity, accessibility, or compliance everywhere else.
Federal anti-discrimination law, EEOC technical assistance, and New York City Local Law 144 have different coverage and legal effect. A bias audit required by one jurisdiction does not establish validity, accessibility, or compliance everywhere else.
U.S. EQUAL EMPLOYMENT OPPORTUNITY COMMISSION
Selection procedures must be valid for their use.
Neutral tests can create unlawful disparate impact unless the employer establishes the required job relationship and business necessity. Primary source →
EEOC AND U.S. DEPARTMENT OF JUSTICE · 2022
Automation must accommodate disability.
The EEOC and DOJ identify risks when software screens out qualified disabled people, creates prohibited inquiries, or lacks a reasonable-accommodation process. Primary source →
NEW YORK CITY DCWP · LOCAL LAW 144
Some automated hiring tools require audit and notice.
New York City prohibits covered use without a recent bias audit, publication of specified results, and required notice to candidates or employees. Primary source →
U.S. EQUAL EMPLOYMENT OPPORTUNITY COMMISSION · 2024–2028
AI-enabled barriers are an enforcement priority.
The EEOC identifies technology used to target, recruit, or make hiring decisions that excludes or adversely affects protected groups as a strategic priority. Primary source →
04 · VERIFICATION
What must be established before output becomes consequence?
An employment model must be verified against the actual job, applicant pool, worker population, and consequence. A vendor benchmark cannot establish that a local deployment is lawful or fair.
ValidityDoes the procedure measure something demonstrably related to the job?
ImpactWho is advanced, delayed, screened out, disciplined, or denied opportunity?
ChallengeCan a person obtain accommodation, correction, and a real human decision?
| VERIFICATION LAYER | THE QUESTION | REQUIRED EVIDENCE | FAILURE IF OMITTED |
|---|---|---|---|
| Job relationship | What job requirement does the signal measure and what evidence validates that relationship? | Job analysis, criterion evidence, versioned assessment, and role-specific validation. | A convenient proxy becomes an unjustified gate to employment. |
| Adverse impact | How do selection and outcome rates differ across protected groups and intersectional populations? | Applicant-flow data, cohort definition, statistical analysis, and less-discriminatory alternatives. | Average performance hides concentrated exclusion. |
| Accessibility and accommodation | Can disabled candidates and workers use the procedure without being unfairly screened out? | Accessibility testing, accommodation notice, alternative process, and response records. | The tool mistakes disability-related interaction for inability to perform the job. |
| Decision and redress | Who reviews the result and how can a person contest incorrect data or inference? | Notice, data access, named reviewer, decision record, correction path, and retention limits. | A consequential decision becomes unreviewable because the model and employer point to each other. |
Operational rule: No automated employment procedure should affect opportunity without role-specific validity, impact monitoring, accommodation, notice, and a human reviewer empowered to change the result.
05 · CONSEQUENCE TEST
Follow the burden to the person or system that carries it.
A concrete pathway reveals where a nominally advisory system becomes practically decisive.
HYPOTHETICAL · PROMOTION
The score is consistent. The opportunity is not.
A company ranks employees for promotion using communication activity, project velocity, manager feedback, and office-presence signals.
01 · DATA
The system treats lower message volume and fewer office days as weak engagement.
02 · CONTEXT
A high-performing employee uses an approved disability accommodation and works asynchronously.
03 · RANKING
The model lowers the employee below the promotion-review threshold.
04 · REVIEW
Managers never see the candidate because the ranking is treated as neutral pre-screening.
No manager intended to discriminate. The procedure converted an accommodation into evidence against advancement.EXPLORE RELATED RECORDS →
SYNTHETIC OUTLAW OBSERVATORY
See the evidence.
The Observatory tracks documented events involving hiring, worker scoring, surveillance, scheduling, platform labor, management automation, and employment accountability.
LOADING LIVE LABOR & EMPLOYMENT RECORDS…
06 · LEADERSHIP TEST
Questions leaders must be able to answer.
Leaders need one register of every automated procedure that can affect recruitment, hiring, pay, scheduling, promotion, discipline, or termination.
Where does AI determine who is seen and who disappears?
Map targeting, ranking, thresholds, queues, and defaults before the formal decision.
What validates the procedure for this job?
General model accuracy is not evidence of job-relatedness.
Which groups are excluded or burdened at each stage?
Measure selection and outcomes, not only completed hires.
How are disability and accommodation handled?
Alternative routes must be real, timely, and non-punitive.
Can a manager and affected worker reconstruct the decision?
Preserve inputs, version, reason, reviewer, correction, and final action.
07 · CONTROL PRIORITIES
What an institution should require now.
Controls must reflect the actor, authority, system, population, data, consequence, and environment of failure.
01 · INVENTORY
Treat AI as an employment procedure.
Register every tool, model, threshold, and automated inference that affects work.
02 · VALIDATION
Prove relevance to the actual job.
Document job analysis, criterion relationship, limits, and revalidation triggers.
03 · IMPACT
Monitor the full funnel.
Track exposure, application, screening, interview, offer, promotion, discipline, and exit outcomes.
04 · ACCESS
Design accommodation into the process.
Give notice, provide accessible alternatives, and prevent accommodation data from becoming a negative signal.
05 · HUMAN REVIEW
Give reviewers authority and context.
Show reasons and uncertainty; prohibit rubber-stamp approval and review only after the decision.
06 · RECORD
Make the process auditable.
Retain versions, data lineage, notices, audit scope, results, overrides, complaints, and corrections.
LABOR & EMPLOYMENT AI EXPOSURE REVIEW
Bring one employment decision into the room.
A focused review maps where one deployed or proposed system changes access to work, pay, scheduling, evaluation, promotion, discipline, or exit. The review tests validity, adverse impact, accommodation, responsibility, notice, and challenge against the real decision path.
- 0190-MINUTE PEOPLE, LEGAL AND EXECUTIVE SESSION
- 02ONE CONSEQUENTIAL EMPLOYMENT PROCEDURE
- 03VALIDITY, IMPACT AND DECISION-PATH REVIEW
- 04WRITTEN VERIFICATION-BURDEN MEMORANDUM
- 05PRIORITIZED WORKER AND EMPLOYER CONTROLS
08 · SOURCES
Primary sources.
This brief relies on selected U.S. federal and local employment authorities. It is not legal advice and does not state the law governing every employer, worker, jurisdiction, bargaining agreement, or employment practice.
U.S. EQUAL EMPLOYMENT OPPORTUNITY COMMISSIONEmployment Tests and Selection Procedures
EEOC AND U.S. DEPARTMENT OF JUSTICE · 2022AI, Software and Disability Discrimination in Employment
NEW YORK CITY DCWP · LOCAL LAW 144Automated Employment Decision Tools
U.S. EQUAL EMPLOYMENT OPPORTUNITY COMMISSION · 2024–2028Strategic Enforcement Plan
SYNTHETIC OUTLAW OBSERVATORYRelated Labor & Employment Records
RECOMMENDED CITATION
Synthetic Outlaw Research. “AI at work: hiring, surveillance, and the right to challenge” Institutional Risk Brief 04, version 1.0. July 21, 2026. https://www.syntheticoutlaw.com/industries/employment.html.