Connecticut May Show Rejected Job Applicants How AI Screened Them
IE 11 Not Supported
For optimal browsing, we recommend Chrome, Firefox or Safari browsers.
A recent law requiring prospective employers to disclose use of AI in the job-screening process is likely to be expanded to allow rejected applicants a chance to see errors made in reviews.
(TNS) — Connecticut’s recent law requiring prospective employers to disclose their use of artificial intelligence in the job-screening process is likely be expanded to allow rejected applicants opportunities to see whether there were errors made in their reviews.
That is the assessment of the two co-chairmen of the legislative General Law Committee as they and a working group of professionals develop potential ideas for the next session of the General Assembly.
“Corporations, or even your local small business, if they’re going to use AI to determine whether you’re going to get fired or hired or get a raise, you should have access to the data that they used in that AI program,” said state Rep. Roland Lemar, D-New Haven. “You should know what information they have that would make you ineligible for the job position.”
Lemar and state Sen. James Maroney, D-Milford, the current committee co-chairmen, also co-chair the state’s working group on AI, which heard suggestions from state and national experts on Monday as Connecticut plans on further AI protections for consumers and in the workplace.
One suggestion builds on a law passed this year requiring notification of job applicants whose applications are screened by AI models.
During a two-hour meeting in the Legislative Office Building in preparation for the upcoming 22-week legislative session starting Jan. 6, the lawmakers said that simply notifying job applicants that they’ll be reviewed by AI is not enough. Job seekers should also be able to look at and possibly correct erroneous conclusions drawn by the nascent technology.
Possible AI review errors include artificial intelligence’s failure to recognize an applicant’s education level, or misinterpreting an applicant’s previous resignation as a firing.
“We just notified people that hey, AI is being used,” Lemar said. “There should be a follow-up at some point, when an adverse decision is made. If you were a job applicant who was discouraged, it would be to check to see whether they got you at the right college, or the right credentials, or whether that was something they just misfiled in their system.”
This year, 23 states approved a variety of new laws governing artificial intelligence, said Tatiana Rice, senior legislative director for the Washington-based nonprofit Future of Privacy Foundation. In remote testimony, she said that new key areas are automated decision making and employee monitoring. Rice noted that anti-discrimination laws in various states need to designate the responsibilities of job-screening AI developers and the companies that use the technology.
“Resume screeners — I think I had read this — that are used in about 90% of Fortune 500 companies, use AI in varying degrees in order to determine whether somebody is fit for a position, whether they get interviewed, whether they get hired as well as internal use of automated systems in order to determine things like compensation; who may be eligible for promotions, et cetera,” Rice said, stressing the need for humans to have the last word on reviewing applicants.
“States have a very productive role to play in not only trying to protect your own constituents, trying to respond to constituents in a way that is both consumer-protective but also implementable by a company,” Rice said, adding that the states are the testing grounds for whatever may emerge for eventual AI legislation on a national basis.
“I think we’d like to get towards post-decision notice and then perhaps some form of the explanation, within a right-to-correct data, if it’s third-party data,” said Maroney, who in recent years has become a nationally known legislative leader on artificial intelligence and online safety.
Crystal Weise, policy and programs director for the nonprofit AFL-CIO Tech Institute in Connecticut, stressed the need for meaningful human oversight that should not be inundated with AI-based decisions that are too numerous to give “thoughtful oversight and review.”
A potential pitfall is so-called confirmation bias that could get humans to go along with possibly flawed employment recommendations made by AI systems, Weise said.
“Some of the provisions require things like the person that’s doing the reviewing, doing the oversight, needs to have training on how the system works, that they’re aware of where it might go wrong,” Weise said. “You know, these systems are not perfect. The person needs to have expertise in that field.”
© 2026 The Register Citizen, Torrington, Conn. Distributed by Tribune Content Agency, LLC.
Never miss a story with the GovTech Today newsletter.