AI hiring tools you can put in front of a regulator
Screening and hiring-decision AI is classified high-risk under the EU AI Act and bias-audited under NYC Local Law 144. EvalGuard ships the bias scorers, candidate-PII firewall, and evidence-bundle audit trail that turn 'trust us' into a defensible file.
What ships today
Honest posture, not roadmap promises
Every checked item is in production today. In-progress items are flagged explicitly — no overclaiming, no vapor.
Built for buyer reality
HR & recruiting AI use cases we ship for
Resume screening + candidate ranking
AI scores inbound resumes against a job profile. Any systematic disparity across protected classes is an adverse-impact finding — you need disparity testing before deploy and a decision trail after it.
EvalGuard features
- Bias scorers run disparity checks across name, gender, and demographic proxies
- Red-team packs run counterfactual bias probes (name, gender, age swaps) against the ranker pre-deploy
- Tamper-evident audit log records model, prompt version + score per candidate
- Evidence-bundle export packages the testing file for your employment counsel
Candidate-facing screening chatbot
A chatbot collects experience and eligibility answers from applicants. It must never ask prohibited questions (age, family status, disability) and must keep candidate PII out of third-party model logs.
EvalGuard features
- Configurable topic guardrails keep prohibited questions (age, family status, disability) out of bot responses
- Candidate-PII firewall redacts name / contact / DOB before the LLM call
- Prompt-injection defense: 300+ attack plugins cover extraction of other candidates' data
- Consent gates at /api/v1/privacy/consent honor candidate opt-outs
Interview summarization + structured feedback
AI turns interview transcripts into structured scorecards. Summaries must be faithful to what was actually said — an invented weakness in a scorecard is a lawsuit exhibit.
EvalGuard features
- Faithfulness scorer checks every scorecard claim against the transcript
- Toxicity + bias scorers flag subjective or protected-class-correlated language
- Per-project BYOK keys keep interview data inside your provider tenancy
- Audit log ties each scorecard to transcript + model version for review
Internal HR policy copilot
Employees ask an assistant about leave, benefits, and disciplinary policy. Wrong answers about legal entitlements create liability; the copilot must ground every answer in current policy documents.
EvalGuard features
- Citation scorer requires policy-document grounding for every answer
- Hallucination gate blocks answers that drift from the source handbook
- Gateway cost ledger + per-tenant budget caps for predictable HR-ops spend
- Multi-model routing serves routine questions on cheaper models within the accuracy threshold
Wire it in 60 seconds
Wrap your OpenAI client. Get a defensible hiring file.
Bias thresholds + PII rules + decision-trail retention are configured once in the EvalGuard control plane. Your code only wraps the client.
import OpenAI from "openai";
import { wrapOpenAI, EvalGuardViolationError } from "@evalguard/openai";
const openai = wrapOpenAI(new OpenAI(), {
apiKey: process.env.EVALGUARD_API_KEY!,
projectId: "resume-screener",
metadata: { vertical: "hr", euAiActHighRisk: true },
blockOnViolation: true, // refuse PII-leaking prompts
evalOnResponse: { failOnScore: 0.7 }, // bias + faithfulness gate
onViolation: (r) => alertPeopleOps(r.violations),
});
try {
await openai.chat.completions.create({
model: "gpt-4o",
messages: [{ role: "user", content: screeningPrompt }],
});
} catch (err) {
if (err instanceof EvalGuardViolationError) {
// Block recorded in the decision audit trail. Replay via audit ID.
}
}wrapOpenAI for wrapAnthropic.Stack
Six surfaces, one platform
Eval, firewall, red-team, audit, BYOK, dashboard — every surface ships out of the box. No bolt-on vendors, no procurement cycle per capability.
Ready to ship hiring AI you can defend?
Free trial includes the bias scorers, candidate-PII firewall, and the decision audit trail. Bring your employment counsel — the evidence bundle is built for them.
Apache-2.0 source · SOC 2 evidence engine live · full trust center