The war between recruiters and job hunters is escalating rapidly due to the use of large language models (LLMs). Candidates have long complained about automated scoring by applicant tracking systems (ATS), which is most noticeable when you get a rejection email just a few seconds after completing a lengthy online application. Thus, those candidates are now having LLMs compare the job description to a resume and then providing an updated resume that includes the keywords that should result in a higher “score” from the ATS. There are new companies every week that promise to get your resume noticed and to get you short-listed for jobs that you never even saw online. Recruiters are now upset because they are getting flooded with applications from people who are not really qualified for the job.

Enter the companies that want to use LLMs and machine learning to help recruiters detect resumes that have been enhanced by AI. And on and on and on…

This tech-fueled war will not end soon, but there are several legislators and lawyers who are starting to pay attention. Let’s review some of the problems companies are about to face:

• Someone is going to do the math and realize that the algorithms (filters) being added to the applicant tracking systems are (inadvertently) discriminating against protected groups. The ADA is also going to trigger liability as these systems will likely appear to help companies avoid needing reasonable accommodations by screening out disabled people.
• Companies in several states (and companies with remote employees in these states) are going to have to actively assess, report, and fix adverse impact; implement processes for managing consent and notices to applicants; and/or publish annual audits that disclose problems. Employment decisions in Illinois, New York City, and Colorado are the first to make companies question the value (and cost) of using AI.

Keep in mind that the vendors who are hawking these LLM-ATS integrations are not the target for the labor lawyers who stand to profit from violations. The companies that use discriminatory tools are the deep pockets that the lawyers will go after. And, once the legal teams identify a system that accidentally discriminates or does not include the required processes and reports, they will sue to get a list of that ATS vendor’s clients. It will be a blood bath.

If your company wants to jump on the bandwagon and leverage an AI-based ATS, make sure you define the functionality based on the most restrictive requirements defined in both current and proposed legislation in every state and territory, county/parish, and city where you might end up with an employee or contractor. Force vendors to prove that their product is free of bias and adverse impact, complies completely with your AI use policy, and produces the reports that will likely be required by current and future regulations. Have clear red lines in your vendor selection process and do not get swayed by cool bells and whistles. The financial and reputational risks are simply too high.