EU AI Act
High-risk classification for hiring AI.
Hiring AI is explicitly listed as high-risk under the AI Act, requiring transparency, traceability, human oversight, and bias risk management. Compliance is not optional in EU markets.
Explainable AI resume screening
AI resume screening that explains every verdict.
The EU AI Act, NYC Local Law 144 (AEDT), and Colorado SB 24-205 all raise the bar on opaque hiring AI. ShortlistTable is designed around the principles those laws codify: every AI verdict is traceable to a screening criterion, a source sentence, and a recruiter override decision — with a per-cell audit trail you can export for your own compliance review. Customers remain responsible for compliance in their own jurisdictions.
Designed around EU AI Act + AEDT principles · Exportable per-cell audit trail · No automated employment decisions
shortlisttable.app / candidate / alice-chen
Candidate
Alice Chen
Backend platform engineer · 7y
Criterion
Has the candidate run Kafka in production for ≥ 12 months?
Yes · AI + recruiter signed offverdict · source-backed
Evidence in resume
“Built and operated Kafka streaming pipelines at Ledger from 2022-present. Owned primary on-call.”
Resume p.1 · Experience · AI verdict 0.91 · recruiter agreed at 2026-05-25 11:02
Every verdict in ShortlistTable links to the exact sentence in the resume that supports it. If we picked the wrong sentence, you adjust the column. If the sentence is right but the verdict is wrong, you override the cell.
Per-cell
Reasoning trace
Exportable
Audit log
Human-in-the-loop
By architecture
0
Automated employment decisions
Why opaque AI is now a legal problem
AI hiring tools are regulated now.
In the last 24 months, three of the largest hiring markets on earth introduced binding regulation on AI-assisted hiring: the EU AI Act classifies hiring AI as high-risk and requires explainability, bias audits, and human oversight; New York City Local Law 144 (AEDT) requires bias audits and candidate disclosure for any automated employment decision tool; Colorado SB 24-205 imposes similar requirements at state level. More are in flight in California, Illinois, and the UK.
All of these regulations have one technical implication in common: if your AI screener is a black box that produces a single score and silently rejects candidates below threshold, you have to either disclose extensively, run periodic bias audits, or both. We took the simpler route: design the product so this regulation isn’t a retrofit.
NYC AEDT (Local Law 144)
Bias audits and candidate disclosure.
Any “automated employment decision tool” substantially used to screen NYC candidates requires an annual bias audit and disclosure to candidates. Tools that don’t expose their inputs cannot produce a meaningful audit.
Colorado SB 24-205
Algorithmic impact assessments by 2026.
Colorado mirrors much of the EU AI Act for high-risk algorithmic systems including hiring. Vendors and deployers both carry obligations.
How explainability is built in
Explainability that is structural, not bolted-on.
01 · Feature
Per-criterion reasoning
Every AI verdict targets exactly one screening criterion. There is no composite mystery score to reverse-engineer — the unit of explanation is the cell.
Criterion 1 · Yes · 0.91Criterion 2 · Review · 0.62Criterion 3 · No · 0.18
02 · Feature
Source sentence per verdict
Every AI verdict cites the exact sentence in the resume that produced it. Auditable to the page, line, and section.
Resume p.1 · Experience“Owned primary on-call for the payments event bus across 14 services.”
03 · Feature
Human override is logged
When a recruiter overrides a verdict, the original AI value, the override, the actor, the timestamp, and the optional note are all retained. The decision is the recruiter’s; the trail proves it.
AI: review · 0.62
Recruiter override · yes · 11:02
04 · Feature
No automated employment decision
There is no path in the system that removes a candidate from consideration without a recruiter action. The bottom of the list is a hold queue, not an auto-rejection. Under AEDT this means we are not, on our own, an AEDT.
No auto-rejectHold queueRecruiter signs offReviewable rejection
05 · Feature
Exportable audit log
Per-workspace audit logs export to CSV / XLSX for periodic bias audits, candidate disclosure requirements, or internal compliance review. Timestamps, actors, and rationale included.
AI verdict · timestamp · confidence
Recruiter override · actor · note
Re-run after column edit · timestamp
06 · Feature
Workspace retention controls
Per-workspace data retention policies — auto-archive after N days, hard-delete after M days — to meet jurisdiction-specific candidate-data rules.
Auto-archive 30dHard-delete 90dPer-workspace policy
How it compares
Explainable AI vs the alternatives.
| Property | Black-box AI screener | ATS rank column | Manual screening | ShortlistTable |
|---|---|---|---|---|
| Per-criterion reasoning | ✕Composite only | ✕Single rank | ✓Manual notes | ✓Per cell |
| Source citation | ✕Rare | ✕No | –Manual | ✓Every verdict |
| Override logged | ✕Edit-only | –Limited | ✓Manual log | ✓Per-cell trail |
| No automated decision | ✕Often auto-reject | ✕Auto-rank cut | ✓By definition | ✓Hold queue only |
| Exportable audit log | ✕Rare | –ATS-shaped | ✕None | ✓CSV / XLSX |
| Per-workspace retention | –Global | –ATS-bound | ✕Manual | ✓Per-workspace policy |
✓ Supported– Partial / manual✕ Not supported
FAQ
Common questions on AI hiring compliance.
Is ShortlistTable an “Automated Employment Decision Tool” under NYC AEDT?+
Our position is no — AEDT applies to tools that substantially assist or replace discretionary decision-making, and ShortlistTable does not auto-reject, does not produce ranked output as the final decision, and requires recruiter action for every screen-out. The recruiter is the decision-maker; we are evidence-gathering with a paper trail. That said, AEDT compliance is the deployer’s responsibility, not the vendor’s. If you use ShortlistTable in NYC, you should still run your own annual bias audit per the law and consult counsel on whether your specific configuration triggers AEDT — we provide the per-workspace audit logs that audit needs.
Are you compliant with the EU AI Act?+
Compliance with the AI Act is a property of how a customer deploys an AI system, not of the vendor alone. ShortlistTable is designed around the Act’s requirements for high-risk hiring AI — per-cell traceability, exportable audit logs, mandatory human override, no automated decision-making — but we do not claim certification (no certification scheme exists yet for hiring AI under the Act). We will publish a conformity-assessment summary as the Act phases in through 2026-27 and update this section accordingly.
Do you train your models on customer candidate data?+
No. Candidate data from your workspace is never used to train any model. We use evaluation-driven model selection on synthetic and licensed datasets, not customer data.
What does the exportable audit log contain?+
Per-cell: criterion, AI verdict, confidence, source citation, recruiter override (if any), actor, timestamp, and free-text override note. Per-workspace: column edit history, re-run history, candidate access log. Exportable as CSV or XLSX for compliance review.
Can a candidate request the reasoning behind a screening decision?+
Yes — the per-cell audit log can be filtered to a single candidate and exported, producing a record of every verdict, source citation, and recruiter action. This supports candidate disclosure requirements under AEDT and similar laws.
How do you handle bias risk management?+
Two layers: (1) at the product level, every verdict surfaces its source sentence, which makes systematic mismatches visible to the recruiter; (2) per-workspace, the override log lets you analyse where the AI verdict and the recruiter disagree systematically — that disagreement is the strongest signal of where bias is hiding, and we surface it explicitly.
Where is candidate data stored?+
Per-workspace, encrypted at rest, with configurable retention (auto-archive after N days, hard-delete after M days). Choice of US or EU region. SOC 2 Type 2 in progress.
More on AI hiring compliance
Keep exploring.
Related
The AI resume screener
Auditable AI screening, end-to-end.
Related
The AI recruiting copilot
Why copilot beats autopilot.
Related
Source-backed screening
Citations on every cell.
Related
Security & compliance brief
Encryption, retention, region.
Read
Keyword vs semantic screening
What changes for compliance.
Built for the regulated era
AI screening designed to hold up under your own audit.
Try ShortlistTable on a 25-resume pile. Every verdict has a source citation, every override has an audit trail, no candidate is silently rejected. Your compliance team gets a paper trail; you stay the decision-maker.