Bias in AI hiring tools isn't a hypothetical concern raised by critics with nothing better to worry about. It's a documented problem with real cases behind it, and it's worth understanding in plain terms rather than either panicking about it or dismissing it.
Where bias actually shows up
The clearest example remains an internal Amazon recruiting tool, scrapped years ago after it was found to be biased against women, having learned its preferences from a decade of resumes submitted mostly by men. In video and voice screening, a different kind of bias shows up in paralinguistic scoring: accents, second-language speech patterns, and some speech disorders can push scores down on delivery even when the underlying answer is strong, since the system is reading surface features of speech alongside content. Bias in AI hiring tools tends to concentrate wherever the system is trained on historical data or judging speech patterns rather than direct, checkable qualifications.
Your rights are starting to catch up
Regulation is starting to address this directly. In New York City, Local Law 144 requires any employer using an automated tool to substantially assist a hiring decision to have an independent bias audit done within the past year, publish a summary of that audit, and give candidates at least 10 business days' notice before the tool is used, along with the right to request an alternative selection process or accommodation. Penalties for violations run from 500 to 1,500 dollars per instance, with enforcement expected to tighten further. Other jurisdictions are moving in similar directions, though coverage still varies a lot by location, and plenty of employers outside covered areas face no comparable requirement yet.
What you can realistically do about it
You generally can't audit a company's algorithm yourself, but you can act on what's within reach. If you're applying for a role covered by a disclosure law, read the notice you're given about the tool and use your right to request an alternative process if something about the format disadvantages you. More broadly, a strong, clearly structured application reduces how much any single scoring quirk can hurt you, since a close match on qualifications and clear, well-organized answers tend to survive imperfect scoring better than a borderline application does. Treat bias as a real factor to be aware of, not a reason to assume every rejection is unfair.