Dear Legislatures, Stop Trying to Figure Out Whether AI “Assists” Employment Decisions

New York Lawmakers Aim to Close Loopholes in NYC’s AI Bias Audit Law and Add Teeth to Workplace Protections

by Amanda Blair, Benjamin Ebbink, and David Walton

at Fisher Phillips

At least two proposed bills pending before the New York State Legislature would force employers to conduct bias audits and provide high levels of transparency if they use AI-fueled automated employment decision tools for employment decisions. This is something that NYC’s first-in-the-nation bias audit law was supposed to accomplish but, in some ways, fell short. These fresh proposals would not only give applicants and employees the right to file lawsuits for alleged violations but would also allow them to sue both employers and the tech developers creating the AI products. What do you need to know about these proposals and how can you prepare?

Key Points From Current Proposals

One of the bills in question sits before the state Senate (S7623) while the other is before the Assembly (A9315). The basics:

• Employers would have to release an annual bias audit report detailing how their use of AI in the employment context held up to anti-bias scrutiny.
• Businesses would also be required to provide a notice to applicants and workers when AI-fueled tools play a role in making employment decisions.
• This would cover all sorts of products that are widely available for purchase and are being heavily marketed to employers by tech vendors, including resume screening tools, AI-related interview products, candidate ranking software, performance assessment tools, and more.
• The bills would also restrict the ability of employers to use AI to surveil and monitor their workers.
• Those aggrieved would have the right to sue – also known as a “private right of action.”
• But they would be able to do more than just sue employers who used the tools. The bills would also allow applicants and employees to file lawsuits against the developers that created the tools and vendors that sold them.

Closing NYC Loophole

New York City’s landmark AI bias audit law was intended to accomplish much of what these state proposals are now hoping to tackle. NYC’s Local Law 144 creates obligations for employers when AI is used for employment purposes, but only when the automated tools play a predominant role in the decisions. As long as employers ensure human managers play that predominant role in the decision-making process, they can properly note that they are not covered by the NYC law.

The state proposals – if passed as currently written – would close that loophole. Human involvement – especially at the end of process after AI has created recommendations that were weighed by decisionmakers (even if not overwhelmingly so) – would no longer absolve an employer from the law’s reach. So long as AI plays a role to assist humans in making employment decisions, those actions would be covered by the proposed laws.

Comparison to California

New York lawmakers are in somewhat of an arms race with the California Legislature, each trying to one-up the other when it comes to building AI regulation in their states. But state legislators hope that the work they do to close the NYC loophole would allow the state to leapfrog ahead of a prominent California proposal.

As noted, the New York proposals would require the transparency notice described above to be sent out whenever AI is used to assist human decision-making. This stands in contrast to a similar California proposal that would require a transparency notification when AI serves as a “controlling factor” in an employment-related decision – a standard that could create another NYC-like loophole.

What’s Next?

The New York State Legislature still needs to work its way through the annual budget process. Once completed in the coming weeks, we’ll have a better sense of the types of bills that can be prioritized by lawmakers.

If the AI bills remain a priority after the budget is addressed, we could see them brought up in committee in April. At that point, we’ll have a much better sense of whether they might actually become law, get watered down by amendments, or get killed by opposition forces before the June legislative deadline.

What Should You Do?

There is a long way to go before these measures ever become law – and any effective date would be in late 2024 or early 2025 – so there is no need to panic and begin wholescale revisions to your policies and practices quite yet. However, there are some steps you might want to consider taking now to put yourself in a good position if we see some movement on the legislative front before the year is up:

• You might want to conduct an internal assessment to see if you are using any AI tools that may qualify under the proposed law. Having a comprehensive inventory will enable you to efficiently address any obligations that come down the road.
• Work with any existing or future AI vendors to understand the steps they have taken to ensure their systems act fairly and without bias.
• Review your vendor contracts carefully to ensure responsibility for any legal mistakes lies with the appropriate party.

Consider a “transparency campaign” well before you are required to. You may find it works to your benefit to inform all stakeholders, including job applicants and employees, about the use of AI in decision-making processes. This might involve creating easy-to-understand guides or FAQs on how AI tools are used, how they can raise concerns, or how to request more information.