So you think AI takes the bias out of hiring? A new study from the National Bureau of Economic Research just found that unchecked AI hiring tools can be biased against certain groups at rates up to 10 times higher than human recruiters. This is a massive problem, and it’s forcing a hard look at how we can actually legislate fairness into these algorithms before they get baked into every company’s recruitment process.
Key Takeaways
- New York City’s Local Law 144, active since January 2023, is already making companies run annual bias audits on their automated hiring tools and publish the results.
- The EU’s AI Act is set to be fully in force by late 2026, and it labels hiring AI as “high-risk,” which means a ton of strict compliance and transparency rules are coming.
- U.S. federal agencies like the Equal Employment Opportunity Commission (EEOC) are putting AI hiring practices under a microscope, hinting that they’ll use existing anti-discrimination laws to bring enforcement actions.
- Companies need to get ahead of this by running their own independent, third-party audits to find and fix bias in their AI systems before regulators come knocking.
- You have to develop internal AI ethics guidelines and actually train your HR teams on what algorithmic fairness means if you want to comply and build a truly equitable hiring process.
“A recent YouGov survey of 19 countries found the UK ranked second-last in the proportion of people who said they had used AI to help them shop. Importantly, 36% of respondents who said they would be shopping around Black Friday didn't plan to use an AI tool at all.”
28% of Companies Still Don’t Audit AI Hiring Tools for Bias Annually
It’s 2026, and a 2025 report from the Society for Human Resource Management (SHRM.org) shows that 28% of companies using AI to recruit are *still* not doing annual bias audits. That figure is alarming. With rules like New York City’s Local Law 144 on the books since early 2023, you’d think this would be standard practice by now. The law is very specific, requiring an independent auditor to check automated employment decision tools (AEDTs) for any disparate impact on candidates based on their sex, race, and ethnicity. If you use AI for screening in NYC, you have to post those audit results right on your website. This goes beyond just checking a compliance box. It’s a basic ethical responsibility. Ignoring this step opens your company up to lawsuits and just keeps old inequalities running on a new operating system. This lack of auditing suggests people either don’t understand the risk or, frankly, don’t care about fairness when they’re chasing efficiency.
The EU AI Act Designates Hiring AI as “High-Risk,” Imposing Stricter Rules by 2026
Across the Atlantic, they aren’t messing around. The European Union’s AI Act, which got its final approval in 2024 and should be fully enforceable by late 2026, puts any AI system for recruitment squarely in the “high-risk” category. This label has huge consequences for anyone building or using these technologies. Under the Act, these high-risk systems have to go through tough conformity assessments before they can even be used, meaning you have to prove you have strong risk management, high accuracy, solid cybersecurity, and real human oversight. And get ready for a lot of paperwork. Mandatory documentation and record-keeping come with transparency rules to give clear info to users. The EU’s goal is to build trust in AI by protecting people’s fundamental rights. For any company hiring from or operating in the EU, following these rules is a prerequisite for doing business. I’m betting this will set the global standard, forcing other countries to step up their game.
Federal Agencies Issue Joint Guidance on AI and Anti-Discrimination Laws
Back in the States, federal agencies are finally coordinating. In May 2025, the U.S. Equal Employment Opportunity Commission (EEOC), the Department of Justice (DOJ), and the Department of Labor (DOL) released joint technical guidance on how existing laws like Title VII of the Civil Rights Act of 1964 and the Americans with Disabilities Act (ADA) apply to AI in employment. The guidance makes it plain that employers are responsible for discriminatory results from their AI tools, even if the bias was unintentional. So, for example, if an AI system screens out more candidates with disabilities because its training data didn’t have enough diverse examples, the employer could be liable. The EEOC has been talking tough about going after algorithmic bias, and this joint guidance shows a unified federal push to apply old laws to new tech. You can’t claim you didn’t know anymore. The burden is now completely on employers to prove their AI tools are fair.
Only 15% of HR Professionals Feel Fully Prepared for AI Bias Regulations
Even with all these new regulations, a 2025 survey by Gartner (Gartner.com) found that only 15% of HR professionals feel ready to handle the complexities of AI bias rules. This readiness gap is a significant risk. The majority said they were only somewhat prepared or not prepared at all. Your HR department is on the front line, using these tools and being the first point of contact for candidates, yet they are often left without the right training or resources to spot, fix, or even explain how an algorithm made a decision. Simply buying an AI hiring tool isn’t a solution. Your HR teams need to understand the tool’s limits, its training data, and its built-in safeguards. Companies have to invest in real training programs on AI ethics, data privacy, and the specific legal rules they’re subject to. Ignoring this training gap will absolutely lead to compliance problems and a damaged reputation.
The Conventional Wisdom: “AI is Inherently Objective” is Dangerously Flawed
Too many people are still clinging to the dangerous myth that AI is naturally objective and magically removes human bias from hiring. The argument goes that because you remove human emotion, you get a pure, data-driven assessment. This perspective shows a fundamental misunderstanding of how AI actually works. These systems learn from historical data, so if your past hiring decisions were biased, and let’s be honest, they probably were, the AI will learn and amplify those same biases at scale. The problem is the flawed data we feed the models and the human choices that guide their development. What if the developers introduced their own biases, or what if the algorithm is picking up on innocent-looking data points that are actually proxies for protected characteristics like race or gender? Relying on AI as a silver bullet without aggressive testing, auditing, and ongoing monitoring is just irresponsible. Real fairness demands constant skepticism and a deep, critical understanding of what this technology can and can’t do. We have to challenge this belief and work to de-bias the entire development pipeline, not just patch the outputs.
The legal ground is shifting fast for AI in hiring, and organizations need to get on top of it now. Engaging with these new rules, from running regular bias audits to properly training HR staff, will help build a more effective and equitable workforce. If you’re digging into how AI affects your employees, it’s worth reading about employee data privacy. As these technologies get woven into our daily work, understanding AI user rights is also becoming essential.
What is algorithmic bias in AI hiring?
It occurs when an AI system produces unfair or discriminatory outcomes against certain demographic groups. This usually happens because the AI was trained on historical data that contained existing societal biases, or because the algorithm itself has flaws that create disparate treatment.
What is New York City’s Local Law 144?
Effective January 1, 2023, this NYC law forces employers who use automated employment decision tools (AEDTs) to get an annual independent bias audit. The results, which check for disparate impact based on sex, race, and ethnicity, have to be posted publicly on the employer’s website.
How does the EU AI Act affect AI in recruitment?
It classifies AI systems for recruitment and employee selection as “high-risk.” This triggers strict requirements for anyone developing or using these tools in the EU, including conformity assessments, risk management systems, human oversight, and transparency duties.
Can employers be held liable for AI bias even if it’s unintentional?
Yes. Federal agencies like the EEOC have made it clear that employers are responsible for discriminatory outcomes from AI tools, even if the bias wasn’t intentional. Existing laws like Title VII of the Civil Rights Act apply to AI-driven hiring decisions.
What steps can companies take to combat AI bias in hiring?
You should be conducting regular, independent third-party bias audits of your AI tools, investing in diverse and clean training data, ensuring strong human oversight, creating internal AI ethics guidelines, and giving your HR teams thorough training on algorithmic fairness and compliance.