This is an opinion column. The facts come from an iCIMS executive’s published essay and two EEOC settlement announcements. The argument is mine: employers who are waiting for AI hiring laws before building governance are waiting for the wrong thing.
A vendor lawyer said the quiet part out loud
On September 29, Courtney Dutter, General Counsel and Chief Compliance Officer at iCIMS, published an essay on five misconceptions about responsible AI in hiring. One of them is that employers need new AI laws before they can act responsibly. She cites iCIMS research finding that 45% of organizations still have no formal AI governance framework, and writes that “waiting for new legislation to address such a gap is waiting for the wrong thing.”
A disclosure is due. iCIMS sells an enterprise talent acquisition platform, and a vendor benefits when its customers feel safe buying AI. I would make the same argument if a regulator had made it, and I think the record supports it.
The law was already there
Dutter’s central claim is simple: “Existing anti-discrimination laws already apply to every hiring decision, including decisions informed by AI.” She points to Title VII, the ADA and the ADEA. None of that is controversial, and it matters for what an HR team should do this quarter. Every screen, score and ranking that touches an applicant is already covered by statutes with decades of enforcement behind them.
What the EEOC’s recent announcements show is how those statutes get applied to a policy an employer improvised in a hurry, and how long the bill takes to arrive. On September 29 the agency announced that Triad National Security, a laboratory management company in Los Alamos, agreed to pay $2.8 million to resolve charges that since at least 2021 it denied religious or disability accommodations to its vaccine mandate. On July 20 it announced a $5 million agreement with an Idaho research and engineering company over a class of more than 100 employees, resolving a 2021 commissioner’s charge and 15 individual charges. In both cases the company resolved the matter without admitting guilt, and the settlement followed a pre-litigation conciliation process. Both require the company to train human resources personnel on accommodation requests.
EEOC Chair Andrea Lucas said of the July settlement: “there was no pandemic exception to workers’ civil rights and liberties.” I would say the same about AI. There is no AI exception. The vaccine policies employers wrote in 2021 were judged in 2026 against laws that had not changed. The AI tools bought this year will be judged the same way, and the gap between a decision and a settlement can be five years.
The counter-argument
The strongest case against me is practical. State and local AI rules differ, some are unsettled, and building a governance program to a moving target can mean building it twice. A cautious buyer can reasonably want the rules fixed before spending on compliance.
I take that seriously, and I still think it fails, for one reason. The work does not depend on which rule wins. Dutter lists what any future law will largely require: documenting processes, testing for disparate impact, maintaining human oversight and being able to explain how decisions are made. Those are the same records an employer would need to defend a decision under Title VII today. Waiting for a statute to tell you to keep them is a choice to have nothing on file when a charge arrives.
What HR should do about it
Dutter’s research also found that 58% of talent acquisition leaders are unclear about the difference between AI and automation. That is the number I would put in front of a chief people officer. If the people running the process cannot say what the tool does, nobody can document it. In the meantime:
- Inventory every tool that screens, scores, ranks or schedules candidates, and write down what each one evaluates.
- Ask each vendor what data the model uses, what bias testing has been done, and where a human can override the output.
- Run disparate impact tests on a fixed schedule and keep the results.
- Name one owner for accommodation requests that arrive through an automated step. I made the case for a task-based approach in an earlier column on full-duty releases, and our coverage of disability suits that begin at the offer stage shows how early in a hiring process an ADA issue can start.
The comfortable reading of the regulatory calendar is that HR has time. I do not think it does. The records a regulator will ask about are made when the tool is chosen, and that has already happened in most companies.
