Amazon Connect Talent vs. bias law
Excellent post from Corey Quinn, which I agree with 100%:
Amazon Connect Talent was just announced. It conducts AI-powered conversational interviews with candidates, generates "anonymized competency scores," and surfaces ranked candidates to recruiters who "make the call."
Fun fact: in New York City, that is an Automated Employment Decision Tool under Local Law 144. AEDTs require an annual independent bias audit with publicly posted results, plus at least ten business days of notice to candidates before use. Illinois, Colorado, and the EU AI Act impose adjacent obligations.
The launch materials mention none of this. The compliance posture appears to be: candidate names are stripped from recruiter dashboards, therefore bias is solved. That is not how any of this works. Proxies for protected class -- speech patterns, zip codes, education history, the resume already sitting in your ATS -- are exactly what bias audits exist to measure.
I don't think the product is bad. I think the announcement is conspicuously missing the guidance customers need before they can deploy it in NYC without violating Local Law 144 on day one.
(The day's other news so far: Amazon Connect now ships as a four-SKU family, and there is a new design philosophy called "humorphism" with its own .com. Both feel small next to the above.)
If you're selling automated hiring decisions in 2026, the bias-audit conversation belongs in the launch.
Tags: bias law amazon aws recruiting regulation automation ai