You’ll see the claim everywhere in HR content: background checks make hiring fairer and more inclusive. It’s a tidy, reassuring idea. In my work across Romania, the UK, I’ve come to a different conclusion: background checks don’t make hiring fair or unfair on their own. They make it whatever your design makes them.
A background check is a tool. Like any tool, it does what the person holding it has designed it to do. Run one way, screening protects a consistent, defensible process and keeps genuinely risky hires out of sensitive roles. Run another way, that same check quietly filters out qualified people, entrenches the bias it was supposed to remove, and lands you in front of a regulator. Same instrument, opposite outcomes. The difference is entirely in the design.
So the honest question isn’t “do background checks help or hurt fairness?” It’s “is yours built well?”
If you would like to listen to the article, here is the AI Podacst:
Why screening exists at all
Before we talk about doing it right, it’s worth remembering why anyone bothers. Candidates misrepresent themselves more often than most hiring managers want to believe. StandOut CV’s research found that 64.2% of employees have lied about skills, experience, or references at least once, up from 55% in 2022. And here’s the part that should sting: roughly 6 in 10 candidates who lie on their resumes land the job anyway, which suggests most screening processes are missing the majority of fraudulent applications.
It’s getting harder, not easier. Employment verification discrepancies rose 44% between 2021 and 2024. So screening isn’t paranoia. There’s a real signal‑to‑noise problem in hiring, and a well‑run check is how you cut through it.
But notice what that justifies. It justifies verification, and it justifies confirming that the person is who they say they are and did what they say they did. What it doesn’t justify, on its own, is blanket rules about who’s allowed through the door. That’s where design starts to matter and where I see most employers drift into risk without realizing it.
The neutral‑tool problem, made concrete
Picture two companies running the exact same criminal‑record check on the exact same candidate.
Company A has a rule: any conviction, any kind, automatic rejection. Quick. Clean. Feels rigorous. In practice, it’s one of the fastest ways to create a discrimination claim, because a blanket exclusion tends to fall harder on some demographic groups than others, and “we apply it to everyone” is no defence when the impact isn’t even.
Company B looks at the same record and asks three questions:
- What was the offence?
- How long ago?
- Does it actually relate to this job?
A decade‑old, unrelated conviction doesn’t block a qualified hire. A recent, directly relevant one gets weighed properly.
Company B isn’t being soft. It’s being legally defensible and better at hiring at the same time. That’s the whole argument in miniature. The tool didn’t change. The design did.
In my experience, most organisations don’t intend to build a “Company A” process. They inherit one: a policy drafted years ago, a hiring manager’s preference that became standard practice, or a screening provider’s default setting that no one revisited.
What good design actually looks like, by jurisdiction
This is where Mindit’s world gets interesting, because “good design” isn’t a vibe. In every market we work in, the law has already written down what a fair process looks like. The guardrails aren’t obstacles. They’re the blueprint.
United States
The backbone here is the EEOC’s guidance on using criminal records, built on what practitioners call the Green factors (from Green v. Missouri Pacific Railroad). The framework calls for individualized assessment considering:
- the nature and gravity of the offence,
- the time that’s passed since it, and
- the nature of the job sought.
A blanket policy excluding everyone with any conviction, regardless of job‑relatedness or time elapsed, creates significant legal risk. The reason is disparate impact under Title VII: blanket exclusions risk violating the law because they produce an uneven effect across groups.
Layered on top is the ban‑the‑box movement. Nationwide, 37 states and over 150 cities and counties have adopted some form of “ban the box,” delaying criminal‑history questions until later in the process, and 15 states extend those fair‑chance rules to private employers. For a company hiring across multiple states, that’s a genuine compliance maze—and a big one: roughly one in three American adults, over 70 million people, has some kind of record.
Adverse action under the FCRA
Then there’s the adverse action process under the FCRA, and this is where a lot of employers trip. If you’re going to reject someone based on a report, you send a pre‑adverse notice, then you wait before finalizing. How long? Here’s the nuance worth getting right: the FCRA itself doesn’t set an exact number of days. It requires a reasonable opportunity for the candidate to respond, and five business days has become the common internal default that the FTC has indicated is reasonable. It’s a widely adopted benchmark, not a statutory rule, and some states add their own longer windows on top. Skip or shorten it and, if the candidate disputes the report, you have to pause entirely until the reporting agency finishes its reinvestigation.
In practice, the companies that get this right treat FCRA adverse action as part of candidate experience, not just a compliance checkbox. They explain what’s happening, they give people a real chance to respond, and they document the decision.
United Kingdom
The UK’s fairness mechanism is the Rehabilitation of Offenders Act 1974, and it changed meaningfully not long ago. From 28 October 2023, under Section 193 of the Police, Crime, Sentencing and Courts Act 2022, the periods after which convictions become “spent” were reduced. In practice:
- custody of six months or less is now spent after one year (down from two),
- custody of more than six months and up to four years after four years (down from seven), and
- custody of more than four years after seven years, which previously could never become spent.
Once a conviction is spent, the candidate can usually treat it as if it never happened. The obvious exception holds: serious violent, sexual, or terrorist offences never become spent and are always disclosed.
One trap for anyone operating UK‑wide: these changes apply to England and Wales only. Scotland and Northern Ireland run their own separate regimes, so a single DBS check won’t cover everyone properly. And which level of DBS check you’re even allowed to run (Basic, Standard, or Enhanced) depends on the role, with spent convictions surfacing only for the exempted positions covered by the ROA Exceptions Order.
In my work, the most common mistake I see here isn’t malicious; it’s lazy. A global policy says “disclose all convictions,” and nobody updates it when the ROA changes. That’s how you end up asking candidates in England and Wales about spent convictions you’re not even allowed to consider.
Romania
For Romanian hiring, the constraint sits in data protection law. Under GDPR Article 10, criminal‑conviction data can only be processed under the control of official authority or where specifically authorized by law with proper safeguards, and Romania’s Law 190/2018 shapes how that plays out locally. On top of that, the Labour Code (Law 53/2003) Article 29 limits recruitment enquiries to what’s genuinely relevant to assessing professional aptitude for the role. You can’t just ask everything because you’d like to know.
There’s a subtle point here that trips up a lot of well‑meaning employers, and it’s worth stating plainly. Getting a candidate’s signed consent is not the same as having a legal basis to process their data. At Mindit, signed consent is always obtained before any check runs, because it’s the right thing operationally and it keeps the process transparent. But that consent functions as documented authorization, not as the sole GDPR legal basis for the processing. Treating a signature as if it settles the legal question is exactly the kind of design flaw that looks fine until someone examines it.
In Romanian practice, I often see companies conflate “we have consent” with “we’re compliant.” They’re not the same. You still need a lawful basis under Article 6, you still need to respect Article 10’s restrictions on criminal data, and you still need to stay within the relevance limit in the Labour Code.
The three side by side
| Aspect | United States | United Kingdom | Romania |
|---|---|---|---|
| Core framework | EEOC guidance / Title VII | Rehabilitation of Offenders Act 1974 | GDPR Art. 10 + Law 190/2018 |
| Key principle | Individualized assessment (Green factors) | Spent convictions after set periods | Lawful basis + relevance to role |
| Criminal‑record limit | No blanket exclusions; job‑relatedness required | Spent convictions treated as if never occurred | Restricted processing; official‑authority control |
| Process safeguard | FCRA adverse action + reasonable response window | Correct DBS level for the role | Labour Code Art. 29 relevance limit |
| Common design failure | Automatic rejection on any record | Asking about spent convictions where barred | Treating consent as the legal basis |
Read that table as a fairness checklist rather than a compliance chore. Every “common design failure” in that last row is a place where a badly built process turns a neutral tool into a discriminatory one. In my experience, these are also the exact points where candidate trust erodes fastest.
Fairness and the business case aren’t in tension
There’s a tempting assumption that doing screening “fairly” means doing it more slowly, more expensively, and at some cost to the quality of your hires. That assumption is wrong, and it’s worth saying so directly.
A poorly designed process leaks candidates. Opaque checks, long unexplained delays, decisions people can’t understand or challenge: all of it drives good applicants to walk before you’ve finished evaluating them. A well‑designed process does the opposite. Consistent criteria mean nobody’s judged on a hiring manager’s mood. A clear adverse action step means a genuine mistake in a report (they happen) gets caught before it costs you a hire you actually wanted. Fair design and good hiring pull in the same direction. The companies that treat screening as a box‑ticking risk exercise tend to get the worst of both: weaker candidate experience and weaker legal protection.
[Client example to go here: a short, anonymized Mindit case where individualized assessment or a corrected report changed an outcome would land harder than any statistic. For instance: “Last year, a candidate for a finance role in Bucharest was initially flagged for a conviction that turned out to be spent under UK rules. Because the process included a proper ROA check and an individualized assessment, we were able to correct the record and proceed with the hire. Without that design, the company would have lost a strong candidate and exposed itself to a potential discrimination claim.”]
The takeaway
Background checks don’t make hiring fair, and they don’t make it unfair. They make it whatever your design makes it. The employers who get real value from screening are the ones who stopped asking whether to check and started asking how, then built a process that verifies what matters, respects what the law protects, and treats candidates like people worth keeping.
At Mindit, that’s the principle every screening programme we run is built around. If you’re reviewing how your checks hold up across Romania, the UK, or the US, I’m glad to talk it through. Reach us at office@mindit.ro or mindit.ro.
