The Senate unanimously approved Thursday a bill aimed at scrutinizing the use of algorithms and artificial intelligence by the Connecticut government to ensure that automated systems are not allowed to make discriminatory decisions.
“The goal is to look at making artificial intelligence more secure and transparent,” Democratic Milford Senator James Maloney said during a brief afternoon debate on the Senate floor. “A lot of questions are: Do we use it? , and I feel it will be clear that we are testing before we use it.”
The bill, which will now be considered by the House of Representatives, requires the Department of Public Service to continue to evaluate the list of agencies using AI and how the technology is being used by state governments. required to be published.
Proponents hope the requirement will reveal where automated systems and algorithms are being used by states. A report released last year by the Media Freedom and Access to Information Clinic found that the state had adopted algorithms in DAS, the Department of Children and Family, and the Department of Education, and that the algorithms were being used to assign students to schools in Hartford. found.
Last month, supporters and members of the United States Civil Rights Commission’s Connecticut Advisory Board called for changes to state freedom of information laws that require government agencies to provide information about their use of algorithms.
Those changes were not included in the proposal passed Thursday, but the bill would require states to develop policies to ensure that the use of automated systems does not create discrimination. Agencies are required to conduct impact assessments prior to system implementation. The bill also bans states from adopting untested AI from next February.
Maloney, co-chair of Congress’ General Law Committee, said the language aims to provide guardrails that prevent automated systems from adopting biases based on user behavior.
“We learn from how the world is now,” Maloney says. “So it accelerates and amplifies our existing prejudices. I would like to confirm.”
If adopted, the proposal would also establish a working group to draft recommendations for a Connecticut AI Bill of Rights, as well as recommendations for future legislation on the use of AI by private companies.
Senators didn’t ask for the opportunity to vote against it because the bill was passed in line with the Senate’s consent calendar. During the debate, Senator Paul Chicarella (R, North Haven) said the bill would give residents insight into how technology is being used by the state.
“Trying to really understand what this artificial intelligence can do is a very complex process,” Cicarella said. “In the end, you’ll basically have transparency that everyone can understand in plain English, not 1’s and 0’s and code.”
Senator Tony Huang (Republican, Fairfield) said Congress faces complex challenges in trying to regulate emerging technologies.
“We are trying to legislate, manage and forecast a rapidly changing and dynamic market,” Huang said. “We are doing our best through due diligence.”
Maloney emphasized that the bill’s purpose is not to ban the use of artificial intelligence or algorithms, but simply to prevent them from having different effects.
“Having great potential [in AI], no doubt,” he said. “We know that AI can have so many benefits, but we also know there are dangers. I’m just there.”
