By Jeff Pole, CEO & Co-Founder of Warden AI
AI is changing how companies hire, but it’s also raising new questions. As adoption accelerates, so does the pressure to show that AI systems are working fairly , transparently, and in line with emerging laws.
That’s why we published the State of AI Bias in Talent Acquisition 2025 , a report based on more than 150 real-world audits and over 1 million test samples.
With contributions from forward-thinking vendors across the TA ecosystem, including SquarePeg, who participated in audits and helped shape key insights, this report offers the most comprehensive look yet at how AI hiring tools are performing across protected groups, and where the real risks and opportunities lie.
The results show a maturing space. Most systems we audited met basic fairness thresholds and some even outperformed traditional human decision-making. But the performance gap between tools is growing.
And as legal and procurement standards catch up, vendors and TA teams alike are rethinking what “good” looks like.
Here are three key trends from the report that every TA leader should be paying attention to:
1. Fairness depends on design, not just intent Across the audits we analyzed, 85% of systems met fairness thresholds (within the “four-fifths rule”) across race and sex.
While our report doesn’t prescribe a single best approach, the variation in performance suggests that design choices, including how candidate data is structured and evaluated, likely play a role in fairness outcomes.
This is an encouraging sign. It suggests that with the right design choices, including behavioral signals, skills analysis, and role-specific scoring, AI can actually reduce bias, not reinforce it.
For teams focused on screening and scoring at scale, the message is pretty clear: fairness starts at the design level.
2. Transparency is emerging as a differentiator One of the biggest surprises in the data we accumulated? Even among tools that passed baseline fairness thresholds, fairness scores varied by up to 40%.
That’s a wide spread, and it’s driving buyer behavior. Increasingly, procurement teams want more than just outputs.
They’re asking for audit results, explainability features, versioned records, and documentation that shows how the system works. They want to see the Trust Marks.
In other words: transparency is no longer a nice-to-have. It’s a product requirement.
3. Compliance is coming faster than expected Laws like NYC Local Law 144 , Colorado SB205, and the EU AI Act are shifting expectations from “future-proofing” to right-now readiness.
But it’s important to remember: bias in hiring has never been legal.
In the U.S., anti-discrimination laws like the Civil Rights Act and the Age Discrimination in Employment Act (ADEA) have long prohibited biased decision-making in hiring, whether those decisions are made by humans or by machines.
AI just makes it more visible and traceable.
Our survey showed that fewer than half of vendors had completed compliance work for even one major AI-specific law.
But the risk landscape extends well beyond AI regulations, and failing to proactively test for bias can open companies to legal exposure under existing employment law.
It is not all doom and gloom. Vendors and TA teams who take a proactive approach are already seeing shorter procurement cycles and stronger stakeholder trust.
One effective way to get ahead is through AI auditing , ideally done monthly and across multiple characteristics to catch issues early.
What talent acquisition leaders can do now The takeaway from our AI Bias report is clear: fairness and compliance are no longer optional, they are core. Here is how talent acquisition leaders can continue to lead the way:
Ask vendors for bias audits: Don’t assume compliance. Request bias audit results and explainability documentation from your vendors.Build transparency into your RFPs: Make transparency a requirement from the start. Include disclosure and audit expectations directly in procurement processes.Champion responsible innovation: Pilot tools that embed transparency and compliance by design, and share those lessons both internally and externally.The path forward AI in hiring is here to stay. But building trust, with candidates, regulators, and customers, means showing your work. That starts with measurable fairness, auditability, and transparency by design.
We’re grateful to all the vendors, like SquarePeg, TA leaders, and practitioners who contributed to this year’s report and whose commitment to transparency and continuous improvement sets a strong example for the industry.
It’s clear that we’re entering a new phase, one where fairness isn’t just about avoiding risk, but about building better systems for everyone.
Read the full State of AI Bias in Talent Acquisition 2025 report
.svg)
.svg)
