Skip to content
HR & Staffing · LL 144 · CO AI Act · EU AI Act

Use ChatGPT and Claude in Hiring — Without the AEDT or AI Act Penalty

PortEden replaces candidate names, protected-class signals, and contact detail with placeholders before your prompt reaches OpenAI or Anthropic. Use any AI for sourcing and screening; the model never sees the underlying candidate.

See pricing

Free for solo recruiters · No credit card · Bias-audit-aligned log built in

Maps to
NYC LL 144
Colorado AI Act
EU AI Act
GDPR
SOC 2
The Risk

What Goes Wrong Without PortEden

You Paste a Resume Stack Into ChatGPT to Rank the Top Five

Those resumes carry candidate names, photos, addresses, ages, and graduation years — every protected-class signal a Title VII plaintiff would subpoena. Sending them to OpenAI without a bias-audited tool, candidate notice, and an impact assessment is exactly the AEDT use NYC LL 144 was written to govern.

Your Sourcer Uses Claude to Draft Cold Outreach From a LinkedIn Pull

The pull holds names, photos, current employers, and years-of-experience markers tightly correlated with age. Claude's context window is now the cleanest copy of a candidate slate that the EU AI Act treats as high-risk processing — retained per Anthropic's policy, not your DPA.

You Connect Outlook to an AI Assistant for Pipeline Productivity

Subject lines like "Smith — passed phone screen, female engineer" leak more than you think. Most AI assistants log thread metadata for as long as their retention allows; one prompt-injection turns the assistant into a disparate-impact discovery exhibit.

How PortEden Protects Your Pipeline

Candidate Identity, Redacted Before It Reaches the Model.

PortEden inspects every field your AI is about to see. Candidate names, photos, addresses, ages, gender markers, and free-text protected-class signals are replaced with placeholders at the boundary — never sent to OpenAI or Anthropic.

Your data
PortEdenRedact
Your AI
Claude
ChatGPT
Copilot
Gemini
Grok
Safe
Sensitive
Redacted
Compliance Reality

What AI Employment Law Actually Requires When Your Team Uses ChatGPT or Claude on Candidates

  • NYC Local Law 144 requires an annual independent bias audit of automated employment decision tools (AEDTs), candidate notice 10 business days before use, and public posting of the audit summary. "We use ChatGPT to rank resumes" sits squarely inside the rule.
  • The Colorado AI Act (effective February 1, 2026) and Illinois HB 3773 (effective January 1, 2026) impose risk-management programs, impact assessments, and applicant notice for high-risk AI in employment — including general-purpose models repurposed for screening.
  • The EU AI Act classifies AI used in recruitment, screening, and evaluation as high-risk under Annex III. Conformity assessments, risk management, logging, human oversight, and post-market monitoring all apply, with fines up to 3% of worldwide turnover (or €15M, whichever is greater) for breaches.
  • Title VII, ADEA, ADA, and EEOC guidance still apply on top. AI vendors that log prompts for training or moderation pull every disparate-impact analysis into a discovery target — including resumes the recruiter ultimately rejected.
NYC LL 144 — Department of Consumer and Worker Protection
The Solution

Built For Recruiters

Candidate-Identity-Aware Redaction in Gmail and Outlook

Replaces 50+ identifier types — candidate names, photos, addresses, DOBs, graduation years, gender markers, and free-text protected-class signals — with placeholders in under 200 ms before the prompt leaves your perimeter. The AI receives only the redacted version; the underlying candidate never reaches OpenAI or Anthropic.

Bias-Audit-Aligned Logging for AEDT Use

Per-requisition record of every prompt, every rank, and every screening decision. The fields LL 144's selection-rate and impact-ratio audit needs — and the EU AI Act's Article 12 logging — drop straight into the audit pack instead of being reconstructed from screenshots.

Per-Jurisdiction Policy Engine — NY, IL, CO, EU

Layer per-state and per-country overrides on your team-wide defaults. NYC LL 144 candidate notice, Illinois HB 3773 impact assessments, Colorado AI Act risk management, and EU AI Act high-risk logging can all run at the same time without a recruiter touching a config.

Works With Claude, ChatGPT, and Copilot — No ATS Replacement

Keep using Greenhouse, Lever, Workday, Bullhorn, or JobAdder. PortEden sits in front of any AI tool — no plugin to install in your ATS, no per-recruiter browser extension. Solo recruiter to staffing firm in days, not months.

Works With What You Already Use
Gmail
Gmail
Inbox-side redaction for AI-drafted candidate replies
Outlook
Outlook
Microsoft 365 inbox-side redaction for sourcer email
Google Calendar
Google Calendar
Strip candidate names from interview meeting titles
Slack
Slack
Strip candidate identifiers from hiring-manager threads
With and Without PortEden

The Same Workflow, Two Very Different Outcomes

Ranking a Resume Stack With ChatGPT
Without
Names, photos, addresses, and ages sent to OpenAI in plain text — an AEDT use without bias audit, candidate notice, or LL 144 audit summary posting.
With
Identifiers and protected-class signals replaced with placeholders before the request leaves your network. The AI ranks on skills; PortEden re-hydrates the names locally for the recruiter.
Drafting Cold Outreach With Claude
Without
Candidate names, photos, current employers, and gender markers sent to Anthropic. Retained per their policy, not your DPA — and pulled into discovery on the next disparate-impact suit.
With
Outreach structure reaches the model; identifiers and gender markers are placeholders. The AI drafts on role and skill match without seeing whose profile.
Inbox Search With Copilot or Gemini During an Open Req
Without
Every matching email — resumes, screen notes, salary history — sent to the AI in plain text. Includes candidates the search ultimately discards, the exact stack a Title VII plaintiff wants.
With
Email content reaches the model with identifiers and protected-class signals replaced by placeholders. The AI ranks results without seeing the underlying candidate.
LL 144 Bias Audit or EU AI Act Conformity Assessment
Without
No structured record of which AI ranked which candidates. The auditor reconstructs the AEDT use from screenshots and inbox archeology.
With
Per-req, per-stage audit log with selection rates, impact ratios, and the prompt-level trail Article 12 of the EU AI Act expects — exportable to your auditor or DPA.
Multi-Jurisdiction Rollout to NY, IL, CO, and EU Teams
Without
Each recruiter follows AI policy by hand; one paste-and-prompt is a multi-state, multi-country compliance problem.
With
Team-wide defaults; per-state, per-country overrides flow from your ATS or HRIS.
Try It on Your Next Req

Five-Minute Setup. Free for Solo Recruiters.

Connect Gmail or Outlook via OAuth. Pick the LL 144 / Colorado AI Act / EU AI Act profile. Keep using ChatGPT or Claude exactly the way you do today — with candidate identity protected by default.

See pricing

Frequently Asked Questions

Does using ChatGPT or Claude with PortEden count as an AEDT under NYC LL 144?
If your AI use "substantially assists or replaces" a hiring decision, it's an AEDT — and that bar is low for resume ranking, screen-question generation, and outreach scoring. PortEden doesn't take you out of LL 144; it gives you the bias-audit-ready log, the candidate-notice workflow, and the structured selection-rate data the rule requires. The annual independent bias audit and the public posting are still on you.
How does PortEden help with the Colorado AI Act and Illinois HB 3773?
Both laws (effective 2026) require risk-management programs, impact assessments, and applicant notice for high-risk AI in employment. PortEden's per-prompt audit log feeds the impact-assessment template, the per-state policy engine emits the right candidate notice, and the redaction layer keeps the underlying candidate detail out of OpenAI and Anthropic's logs in the first place.
What about the EU AI Act for recruitment in the EU?
The EU AI Act classifies recruitment, screening, and evaluation as high-risk (Annex III) and triggers the Article 9–15 obligations: risk management, data governance, technical documentation, automatic event logging, transparency, human oversight, and accuracy/robustness/cybersecurity. PortEden's logging and oversight workflows are designed against Article 12 + 14; the redaction layer addresses Article 10's data-quality and processing-minimisation expectations.
Will PortEden change my ATS workflow?
No. PortEden runs in front of your AI, not your ATS. You keep using Greenhouse, Lever, Workday, Bullhorn, JobAdder, or whatever your team runs on, and the redaction layer sits between any AI tool and the data the AI is about to read.
How does PortEden handle hiring-manager Slack channels and informal screens?
PortEden treats Slack as a redaction surface alongside Gmail and Outlook. Hiring-manager threads with candidate identifiers and free-text impressions get the same treatment as inbox content, and the audit log captures who saw what — closing one of the most common LL 144 selection-rate audit gaps.
Can multi-jurisdiction teams apply different rules per US state and EU country?
Yes. Set team-wide defaults once; per-state and per-country overrides flow from your ATS or HRIS. A New York requisition rides LL 144 + Article 14 oversight rules; a Chicago req picks up HB 3773 + Article 12 logging; a Berlin req runs full EU AI Act high-risk controls — without a recruiter re-configuring anything.
What does it cost and how long does setup take?
There's a free tier for solo recruiters. Team pricing scales by recruiter — full pricing is on the pricing page. Setup is under 5 minutes for a solo recruiter on Gmail or Outlook + ChatGPT or Claude. Multi-jurisdiction teams typically take a half-day for SSO and ATS integration.

Keep Exploring

Ready to Use AI in Hiring Without the LL 144 or AI Act Penalty?

Five-minute setup. Free for solo recruiters. Bias-audit-aligned log and per-jurisdiction policy engine from day one.

See pricing

Global staffing firm or RPO? Talk to sales →