AI-assisted layoffs that weigh sick days and medical leave

Nearly a third of managers who use AI to help decide layoffs told the tool to weigh frequent sick days or medical leave. Fourteen percent had it weigh age. The red light should be flashing here if it isn’t already a four alarm alert.

An algorithm didn’t pick up that bias from old data. A manager typed “age” and “medical leave” into a decision tool on purpose.

Age and medical leave aren’t neutral productivity inputs. They’re the exact categories the ADEA and ADA exist to protect. When someone instructs a model to weigh them, that instruction becomes evidence. A plaintiff’s lawyer won’t need statistical inference. They’ll have an admission sitting in the prompt history, fully discoverable.

This is a governance and training problem, not a technology problem. Every tool can be used incorrectly. A manager who wouldn’t dream of writing “consider her medical leave” in a termination memo will type it into a prompt without a second thought, because it doesn’t feel like a document yet. It is one. Any input that touches an employment decision needs the same review a lawyer would give the memo.

Our job as leaders is to teach the right way. If not, count on people using AI to break the law and let the plaintiffs bar do the teaching.

P.s. don’t fire people based on protected statuses.