Screening can quietly widen the gaps in your workforce or help close them. Getting it right is not only what the law requires, it is a better way to build a team, and most employers have never made that choice on purpose.
Key takeaways
- A background check feels objective, but the records it relies on were produced by a system that already treats groups differently.
- Blanket exclusions on criminal or credit history can create disparate impact liability under Title VII, and intent is not required.
- Removing information, as Ban the Box did, does not remove bias; reform has to be designed, not just declared.
- A fair process rests on individualized assessment, job-relevant criteria, consistent application, and regular audits.
- AI screening does not lower your exposure; a biased model reproduces bias faster, and using AI is not a defense against a Title VII claim.
If you run background checks the way most US employers do, you have a bias problem you have probably never measured. Not because anyone on your team intends to discriminate, but because the records you screen against were produced by a system that treats groups differently long before a candidate reaches you. A background check feels objective. It rarely is.
Start with the arithmetic. Roughly as many Americans have a criminal record as hold a college diploma, close to 70 million people. Black Americans are incarcerated in state prisons at nearly five times the rate of white Americans, and the gap for Latino Americans is wide too. So any policy that treats any record as an automatic disqualifier will screen out people of color at a higher rate, whatever your intent. The law has a name for this. It is disparate impact, and since Griggs v. Duke Power in 1971 you do not need to intend discrimination to be liable for it. The effect is enough.
Credit history tells the same story in a quieter voice. Research does not show that credit scores predict job performance, yet scores track decades of unequal access to credit, so a credit filter can act as a stand-in for race. More than a dozen states have restricted credit checks in hiring for that reason. The word “neutral,” applied to either tool, does not survive contact with the evidence.
None of this is a legal gray area, and it helps to be precise about why. Title VII is clear that a neutral-looking policy with a discriminatory effect can be unlawful, the EEOC’s guidance on background checks is unambiguous that blanket criminal exclusions are not job-related, the FCRA spells out consent, disclosure, and the right to dispute, and more than 37 states have passed Ban the Box laws. The framework is settled. What is not settled is practice. Enforcement is inconsistent and under-resourced, and many employers still run blanket exclusions or apply screening unevenly, usually through inertia rather than any decision to break the law. The gap is not in the statute. It sits between what the law requires and what hiring teams do.
Removing information is not the same as removing bias
This is where the conversation usually gets simplistic, and it is worth resisting that. Reform is not automatically better than the thing it replaces.
Ban the Box was meant to help: take the criminal-history checkbox off the application and candidates get judged on their merits first. But one major study found that in some markets the policy reduced employment for young, low-skilled Black men by about five percent. Denied the actual record, some employers fell back on assumptions about who was likely to have one. Removing the information did not remove the bias. It moved the bias to an earlier, less visible point in the process. The lesson is not that reform fails. It is that reform has to be designed, not just declared.
What a fair process looks like
A screening process reduces bias when it is built to, and the components are well understood.
Start with individualized assessment, the approach the EEOC has recommended since 2012. Instead of an automatic yes or no on a record, you weigh the nature and gravity of the offense, the time elapsed since it, and its relevance to the specific job. That lets a candidate show that a ten-year-old conviction has nothing to do with the role in front of them.
Around it, apply the same criteria to every candidate in a job category, tie each criterion to the actual duties of the role, and stop treating arrest-only records as disqualifying, because an arrest is not proof that anything happened. Then check your own work: break your hiring data down by group and apply the four-fifths rule, so that if any group clears a stage at less than 80 percent of the top group’s rate, you look again at why.
AI does not get you off the hook
Screening software now flags records, scores risk, and recommends outcomes, all sold on consistency. The catch is that a model trained on a history of biased decisions learns to reproduce them, then applies them at a speed no human reviewer could match. The bias does not go away. It gets faster, and it arrives wearing the authority of data.
Colorado’s SB 26-189, signed in May 2026, now requires employers using AI that materially influences a hiring decision to disclose it and to offer a right to human review, and it is explicit that complying is not a defense against a Title VII claim. Using AI to screen does not lower your exposure. Handled carelessly, it raises it.
Where KRESS stands
KRESS supplies human verification and a documented, consistent workflow. The hiring decision, and the individualized assessment behind it, belongs to the employer. A provider that appears to make that call for you is not a convenience; it is a liability. Our job is to give you a defensible process and a clean audit trail, not a verdict.
We believe compliance is the floor, not the goal. A diverse workforce is worth building, and the evidence is strongest where it counts: diverse teams make better decisions, they generate more from innovation, and a wider net reaches more capable people.
However, the popular headline that diverse companies are some precise percentage more profitable is a correlation, not proven cause, and researchers such as Alex Edmans have fairly challenged it. The case for better decisions, stronger innovation, and a deeper talent pool does not rest on that headline. Justice-impacted employees in particular tend to perform as well as or better than their peers, with turnover often lower. The fairness case and the business case are the same case.
The honest conclusion
Most employers did not design their screening to be fair or unfair. They designed it to reduce risk, and bias came along as an unexamined side effect. With the Washington Fair Chance Act live on 1 July 2026 and more states moving each year, leaving that side effect unexamined is now a legal exposure and a competitive disadvantage. The law settled the minimum a long time ago. The real question is whether you treat inclusion as a box to tick or a better way to build a team.
The fixes are knowable: individualized assessment, job-relevant criteria, consistent application, and regular audits. Laws and thresholds vary by state, which is why we keep our state-by-state compliance guide up to date.
Ready to build screening you can stand behind? KRESS makes it consistent, documented, and defensible. Get started.
Screen smarter. Hire better.