Back in 2020, I used to be in the midst of a Workology Podcast series focused on the long run of labor for individuals with disabilities. On the time, “AI in hiring” was still being positioned as the answer to bias—not a possible source of it.
I didn’t desire a press release version of that story. I wanted to listen to directly from the people constructing the technology.
So I invited Steve Feyer, then Director of Product Marketing at Eightfold AI, to hitch me for Episode 224. We spent nearly half-hour talking about conscious and unconscious bias, how algorithms will be designed to “forget” protected characteristics, and what it looks prefer to proactively test for fairness before deploying AI into hiring workflows.
We talked about their Equal Opportunity Algorithm. About statistical validation. About transparency.
And I’ll be honest—I walked away impressed. I said, on the record, that this was precisely the form of considering the HR tech space needed more of.
Fast forward to January 2026.
Eightfold AI is called in a federal class motion lawsuit.
This Isn’t the AI Bias Story You Think It Is
In case you’ve seen the headlines, they likely frame this as one other example of AI gone fallacious or biased algorithms harming candidates.
That’s not actually what this case is about.
The lawsuit—Kistler et al. v. Eightfold AI Inc.—does not claim that Eightfold’s technology produced discriminatory outcomes.
As an alternative, it raises a unique—and arguably more disruptive—query:
Is AI-powered candidate scoring subject to the Fair Credit Reporting Act (FCRA)?
Originally passed in 1970, the FCRA governs how consumer data is collected, used, and shared—primarily within the context of credit reporting and background checks.
The plaintiffs argue that Eightfold crossed into that territory.
In response to the grievance, the platform allegedly aggregated data on multiple billion individuals from public sources—job boards, profession sites, resume databases, census data, and social media. That data was then used to generate candidate scores and rankings, which could filter applicants out before a recruiter ever reviewed them.
The problem isn’t just the scoring.
It’s the lack of disclosure, consent, and candidate rights—including the power to see or dispute the data getting used to guage them.
Eightfold has denied these allegations, stating that its data comes from candidates themselves or from employer-provided sources. The case continues to be in its early stages.
But for HR leaders, the consequence isn’t the one thing that matters.
The signal is.
When Transparency Meets Reality
What keeps coming back to me is that 2020 conversation.
Because transparency was a central theme.
Steve talked about “Algorithmic Transparency” as a core product capability—explaining that each employers and candidates could receive clear insights into how matches were made. He emphasized the importance of testing for bias across gender, age, and ethnicity before deployment.
It was thoughtful. It was measured. And it reflected where the industry was heading.
But there was one query I didn’t ask—and truthfully, one that almost all of us weren’t asking on the time:
Where is all of this data actually coming from?
Looking back, the reply was there. Public data sources. Massive aggregated datasets. Machine learning models generating predictions about candidate success.
What wasn’t clear—and what this lawsuit challenges—is whether or not candidates ever had visibility into that process or control over it.
And that’s the shift we’re seeing now.
This isn’t nearly whether AI is fair.
It’s about whether AI is accountable.
Why This Matters More Than You Think
Here’s the fact HR leaders need to grasp:
If AI influences a hiring decision, you own that call.
It doesn’t matter if the model was built by a vendor.
It doesn’t matter if the info got here from sources you’ve never seen.
It doesn’t matter should you can’t fully explain how the scoring works.
From a legal standpoint, the responsibility still lands with the employer. And that’s where many organizations are exposed.
Not because they intended harm. But because they trusted the technology without fully interrogating it.
The Questions We Should Have Been Asking All Along
During that 2020 interview, Steve actually shared several questions HR teams should ask AI vendors.
They were good questions then. They’re essential questions now, together with a couple of more we’ve learned so as to add.
When evaluating any AI hiring platform, you wish clear answers to:
- How are you testing for bias—and may you show statistical validation, not only methodology?
- What external data sources are used to guage candidates?
- Are candidates aware their data is getting used in this manner?
- Can candidates access, review, or dispute their assessments?
- Does your system rating or rank candidates before human review?
- How do you define your role in relation to FCRA compliance?
- What liability protections—and exposures—exist in our vendor agreement?
- Can we independently audit your outputs?
If those questions feel uncomfortable to ask, that’s the purpose.
Because the chance doesn’t come from asking hard questions.
It comes from not asking them.
From Trust to Due Diligence
In 2020, I trusted what I heard in that conversation. And to be fair, it reflected where the industry was on the time—optimistic, innovation-driven, and focused on solving bias through technology.
In 2026, the expectations are different. Trust isn’t any longer enough.
HR leaders need documentation, transparency, and accountability—not only from their teams, but from every vendor influencing their hiring decisions.
Since the organizations that get pulled into this next wave of AI-related litigation won’t necessarily be those acting recklessly.
They’ll be those who didn’t look closely enough.
Take heed to the Original Conversation
If you ought to hear that 2020 discussion for yourself, Episode 224 of the Workology Podcast—Eliminating Algorithmic Bias in Hiring and Employment—you possibly can listen right here.
It’s still a useful conversation.
But listening to it now, with the whole lot we’ve learned since, hits very in another way.