AI in the workplace: Businesses brace for new rules
States are tightening rules on AI in the workplace, while the federal government pushes for deregulation, setting up a battle over hiring and bias laws.
President Donald Trump has moved to deregulate AI on the federal level, but some states want to regulate the use of AI in the workplace. Those dueling goals might collide as states push for stricter AI regulations while Trump pushes for less interference.
California, New York, Connecticut and Texas are among the states considering bills this year to regulate AI in the workplace.
The various legislative proposals designate any workplace AI system involved in hiring and firing as "high-risk" and subject to special requirements. That could include disclosing the use of AI to job candidates and limiting reliance on automated decision-making in any consequential employment action.
But the test of any new law is how authorities enforce it, said Reena Richtermeyer, an attorney at CM Law. She will be watching for that.
"It's one thing to have legislation -- it's quite another to enforce it," Richtermeyer said. She said some of the legislation might be vague as well as difficult to understand, implement and enforce.
Enforcement will tell more about law
Richtermeyer suggested that the first significant enforcement cases -- at the state or federal level -- will set the tone for how seriously businesses take AI laws.
The focus this year is on getting something passed in a state legislature. In California, for instance, the No Robo Bosses Act introduced March 6 seeks to bar employers from exclusively relying on AI in hiring, promotion and termination decisions without human oversight.
"AI must remain a tool controlled by humans, not the other way around," said State Sen. Jerry McNerney, a Democrat from the San Francisco Bay area community of Pleasanton, in a statement.
Similarly, in new legislation in Connecticut, An Act Concerning AI, also seeks human oversight of AI systems with the potential to make consequential decisions, which includes those made by HR departments.
"The states will take the lead, regardless of what happens at the federal level," said David Trier, vice president of product at ModelOp, an AI governance software company.
Trier said businesses must consider the practices these bills seek because they will soon be affected. He said that any company that does business in an AI-regulated state will have to comply with the state's law.
What Trump wants
Meanwhile, in an earlier executive order on AI, Trump called for removing barriers, reducing regulation and eliminating anything that hindered AI development.
The Trump administration asked the public to share their AI policy ideas as part of that effort. The deadline for submitting responses is midnight today.
The Equal Employment Opportunity Commission, under President Joe Biden's administration, made clear that AI could be enforced under existing federal employment discrimination laws.
However, Robert Taylor, an attorney at Carstens, Allen & Gourley, warned that these emerging state AI laws are shifting from the well-defined legal concept of "discrimination" to the more ambiguous concept of "bias" in employment decisions.
"Bias is kind of this squishy concept," Taylor said. Employment discrimination is a defined legal term with a body of case law behind it. But varying definitions of bias across state laws could potentially expose companies to broader liability under easier-to-prove standards than traditional discrimination law would require, he said.
A bias standard might fail to recognize the possibility that AI is improving fairness, Taylor argued. "Just because it's different doesn't mean it's worse," he said of AI-generated results. "It could actually mean it's better if the AI is stripping out the human bias that goes into the decision-making process."
Patrick Thibodeau is an editor at large for Informa TechTarget who covers HCM and ERP technologies. He's worked for more than two decades as an enterprise IT reporter.