Framework · Employment AI
Employment-AI evidence for the AI that screens people.
Illinois HB 3773 has been in force since January 2026, New York City's Local Law 144 since 2023, and Colorado SB 26-189 follows in January 2027. Meilynx produces the runtime evidence for the AI-traffic slice — which hiring-AI systems run, what candidate data reaches them, whether they changed, and a tamper-evident record of AI-assisted decisions — while your independent bias audit stays yours.
Inventory, notify, audit, record.
The operational core across Illinois, New York City, and Colorado is evidential: know which AI systems touch employment decisions, tell the people they screen, prove the tools were independently audited, and be able to show what the AI actually did when an effect claim arrives.
- Discriminatory-effect liability and the zip-code-proxy ban for AI used in recruitment, hiring, promotion, and other employment decisions (Illinois, §2-102(L)(1)).
- Notice to employees and candidates that AI is used, with the timing and content each statute sets (Illinois §2-102(L)(2); NYC §20-871(b)).
- An annual independent bias audit before an automated employment decision tool is used, with a published summary of results (NYC §20-871(a)).
- Disclosure readiness for automated decision-making technology in consequential employment decisions from 1 January 2027 (Colorado SB 26-189).
Runtime obligations, to runtime evidence.
A specific Meilynx control for each obligation the proxy can substantiate, and the artifact it produces. The bias audit, the notices, the scoping determination, and the adverse-impact analysis stay attested in your own program — and the package says so per control.
Employment-AI statutes → Meilynx controls (AI-traffic slice)
| Requirement | How Meilynx maps | Examination artifact |
|---|---|---|
Know which AI systems and automated employment decision tools are in use NYC LL 144 · §20-871(a) · IL HB 3773 · §2-102(L)(2) | The inventory of hiring-AI models and providers is derived from live traffic and enforced by runtime model allow-listing — the set of tools that must carry a current bias audit and be named in candidate notices starts from what actually runs. | Hiring-AI system inventory, generated from traffic |
Keep a record of AI-assisted employment decisions that survives a discriminatory-effect review IL HB 3773 · §2-102(L)(1) | Every hiring-AI request, response, and governance decision is captured inline at the proxy and sealed into a tamper-evident, hash-chained record — the evidentiary substrate when an effect claim asks what the AI was actually asked and answered. | Hash-chained decision audit trail for the period |
Know what candidate data is collected and sent to the tool NYC LL 144 · §20-871(b) | Personal-data detection over hiring-AI traffic evidences the types of candidate data actually flowing to each tool — the fact base behind the data-type, source, and retention disclosure. | Candidate personal-data findings by tool |
Detect unapproved changes to the system as deployed IL HB 3773 · §2-102(L)(1) | Prompt-drift and tool-grant-drift detection flag unapproved changes to hiring-AI configuration against approved baselines — liability attaches to the system as it runs, not as it was designed. | Drift findings against approved baselines |
Evidence the independent bias audit, its publication, and the adverse-impact analysis NYC LL 144 · §20-871(a) · 29 CFR §1607.4 | The bias audit is performed by an independent auditor by statute — Meilynx never performs one. The package carries the auditor, audit date, distribution date, published-results URL, and your four-fifths analysis as attested evidence next to the runtime record. | Bias-audit log and notice evidence, attested |
Know which AI systems and automated employment decision tools are in use
NYC LL 144 · §20-871(a) · IL HB 3773 · §2-102(L)(2)
Maps to · The inventory of hiring-AI models and providers is derived from live traffic and enforced by runtime model allow-listing — the set of tools that must carry a current bias audit and be named in candidate notices starts from what actually runs.
Examination artifact · Hiring-AI system inventory, generated from traffic
Keep a record of AI-assisted employment decisions that survives a discriminatory-effect review
IL HB 3773 · §2-102(L)(1)
Maps to · Every hiring-AI request, response, and governance decision is captured inline at the proxy and sealed into a tamper-evident, hash-chained record — the evidentiary substrate when an effect claim asks what the AI was actually asked and answered.
Examination artifact · Hash-chained decision audit trail for the period
Know what candidate data is collected and sent to the tool
NYC LL 144 · §20-871(b)
Maps to · Personal-data detection over hiring-AI traffic evidences the types of candidate data actually flowing to each tool — the fact base behind the data-type, source, and retention disclosure.
Examination artifact · Candidate personal-data findings by tool
Detect unapproved changes to the system as deployed
IL HB 3773 · §2-102(L)(1)
Maps to · Prompt-drift and tool-grant-drift detection flag unapproved changes to hiring-AI configuration against approved baselines — liability attaches to the system as it runs, not as it was designed.
Examination artifact · Drift findings against approved baselines
Evidence the independent bias audit, its publication, and the adverse-impact analysis
NYC LL 144 · §20-871(a) · 29 CFR §1607.4
Maps to · The bias audit is performed by an independent auditor by statute — Meilynx never performs one. The package carries the auditor, audit date, distribution date, published-results URL, and your four-fifths analysis as attested evidence next to the runtime record.
Examination artifact · Bias-audit log and notice evidence, attested
What you show an agency or a plaintiff's counsel.
The audit trail renders into a compliance package scoped to the AI-traffic slice — every control classified as proxy-verified runtime evidence or attested in your program, with an obligation timeline that names which statutes are in force and which are readiness.
In the package
- Hiring-AI system inventory, auto-populated from traffic.
- Bias-audit log: auditor, audit date, distribution date, published URL — attested.
- Notice evidence and the systems each notice must cover.
- Tamper-evident decision audit trail across the reporting period.
- Obligation timeline — Illinois and NYC in force, Colorado readiness, EU AI Act pointer.
Employment AI and the proxy.
Which employment-AI laws does the preset cover?
Four, under one preset: Illinois HB 3773 (775 ILCS 5/2-102(L), in force since 1 January 2026 — discriminatory-effect liability, the zip-code-proxy ban, and the notice duty), New York City Local Law 144 (enforced since 5 July 2023 — annual independent bias audit, published summary, candidate notice), Colorado SB 26-189 (signed May 2026, effective 1 January 2027 — covered as readiness only), and the EEOC Uniform Guidelines' adverse-impact record-keeping. EU AI Act Annex III employment obligations stay with the EU AI Act preset.
Does Meilynx perform the NYC bias audit?
No — and it must not. Local Law 144 requires an impartial evaluation by an independent auditor, performed no more than a year before the tool is used. Meilynx evidences that the audit exists and is current, alongside the runtime record of which tools actually ran. The audit itself, its publication, and your adverse-impact analysis are attested in the package, and each control is explicitly classified as proxy-verified or attested in your own program.
Does the preset make us compliant with Illinois HB 3773 or Local Law 144?
No runtime product could. Notices to candidates, the scoping call on which systems are automated employment decision tools, feature exclusion, and the bias audit all happen outside the proxy. What Meilynx substantiates from live traffic — which hiring-AI systems run, what candidate data reaches them, whether their configuration changed, and a tamper-evident record of AI-assisted decisions — carries proxy-verified evidence. Everything else is honestly marked as attested.
What about Colorado?
Colorado's SB 26-189 replaced the state's earlier AI Act and takes effect on 1 January 2027 as a transparency and disclosure regime for automated decision-making technology in consequential decisions, including employment. The preset covers it as readiness — an inventory of the technology in use and the notice, disclosure, correction, and human-review procedures drafted ahead of the effective date. No Colorado obligation is claimed as in force before 2027.
See exactly what an examiner receives
Download a sample examination package — model inventory, control coverage, a governance policy snapshot, and a SHA-256 integrity hash.