Connecticut's AI Adventure: Senate Passes Amended SB 2
Key highlights this week:
We’re currently tracking 582 state bills (plus another 102 congressional bills) related to AI this year.
Governors in Iowa, New York, and Alabama all signed sexual deepfake bills into last.
New York also enacted a political deepfake bill and set up funding for an AI research center as part of the state budget.
And the Connecticut Senate passed that chamber’s comprehensive AI bill while making a series of major amendments, which we detail in this week’s deep dive below.
Events have unfolded quickly in Connecticut, where lawmakers have tweaked a potential landmark AI bill in response to stakeholder concerns. The unlikely state has become the key battleground for AI policy in the United States. With just weeks to go before the May 8 adjournment date, the bill cleared the Senate on Wednesday but still faces an uncertain future in the House as long as Governor Ned Lamont (D) is not fully on board.
It only took a few days for part of last week’s roundup of major comprehensive AI bills to become out-of-date, as Connecticut lawmakers announced further changes to their priority AI legislation (CT SB 2). The Senate passed the amended version of the bill on a 24-12 vote after over four hours of debate, sending the bill to the House for consideration before lawmakers adjourn for the year in less than two weeks.
In general, the bill provides obligations for developers and deployers of AI systems, requiring “reasonable care” to protect against algorithmic discrimination. The biggest change the bill undertook this week was the deletion of a provision requiring developers of general-purpose AI systems to provide certain documentation and disclosures about the models and testing. Instead, the legislation would now create a working group to recommend legislation to regulate the use of general-purpose AI. The amendments also push back several effective dates with many of the obligations on developers and deployers beginning on July 1, 2025, rather than October 1, 2024.
Other changes to the bill include:
Amending the definition of “consequential decision” to no longer include decisions on utilities;
Addition of a definition for “health care service”;
Exemption of anti-fraud and cybersecurity software and certain consumer chatbots from requirements as high-risk AI systems;
Clarification of the definition of “substantial factor”;
Requiring developers of “high risk AI systems” to disclose high-level summaries of the type of data used to train systems;
For requirements that developers make certain disclosures to deployers, exempting developers that are also deployers;
Revising statement requirements deployers must make to consumers;
An exemption for banks if they are in compliance with banking regulations;
Requiring deployers to provide certain information to consumers regarding consequential decisions with a chance to correct inaccuracies in the underlying data;
Additional exemptions from deployer obligations, including for small businesses;
Providing rebuttable presumptions and affirmative defenses only apply to attorney general actions;
Providing for a liaison between legislative leaders and the Connecticut Academy of Science and Engineering, with the development of certain tools to determine compliance;
Deleting provisions creating an AI Advisory Council; provisions requiring agencies to study how generative AI can be used to improve government efficiency, and state employee training on AI;
Expanding the state’s information and telecommunications strategic plan to include generative AI;
Deleting grant programs in the bill; and
Repurposing the Technology Talent Advisory Committee to include AI and quantum computing.
The bill still retains provisions requiring disclosure to consumers interacting with AI, labeling requirements for synthetic content, prohibiting political and sexual deepfakes, and integrating AI with workforce development programs and online courses with certifications.
Governor Ned Lamont (D) has expressed reservations about the regulatory sections of the bill, and House Speaker Matt Ritter (D) has warned he will not call a vote until the governor’s concerns are addressed. Upon passage of the bill, a spokesperson for the governor was noncommittal on the bill in a statement. "We are supportive of efforts to increase disclosure of the use of AI in elections and to close loopholes around intimate images. However, the governor remains concerned that this is a fast-moving space, and that we need to make sure we do this right and don't stymie innovation."
With these amendments, Google dropped its opposition to the bill, and Microsoft and IBM came out in favor of it. IBM even published an opinion column praising the legislation as a “beacon of foresight and balanced regulation” that avoids broad regulation and focuses on high-risk applications.
The changes also come after lawmakers from several states held a press conference last week to discuss the pushback they had received on AI legislation from competing interest groups. “Every bill we run is going to end the world as we know it. That’s a common thread you hear when you run policies,” remarked Colorado Senator Robert Rodriguez (D).
With the Senate passage of SB 2, attention will remain on Connecticut over the last twelve days of the legislative session to see if lawmakers will set a standard of AI regulations for other states, and potentially the federal government, to follow.
Recent Developments
Major Policy Action
Iowa: Last Friday, Gov. Reynolds (R) signed a sexual deepfake bill (IA HF 2240) into law. The new law makes it a crime to harass by the dissemination, publishing, distribution, or posting of a visual depiction showing another person in a state of full or partial nudity or engaged in a sex act that has been altered to falsely depict another person.
New York: On Saturday, Governor Hochul (D) signed the state budget (NY AB 8808 & SB 8308) into law, which includes a right of action against non-consensual sexual deepfakes and prohibits political deepfakes without a disclaimer. The budget also includes the establishment of “Empire AI,” a consortium that will create a state-of-the-art artificial intelligence computing center in Buffalo to be used by New York colleges and universities. The budget will allocate $275 million for Empire AI, and $125 million in private and university funding for artificial intelligence development.
Alabama: on Wednesday, Gov. Ivey (R) signed two sexual deepfake bills into law. The first new law (AL HB 161) makes it unlawful to knowingly create, record, or alter a private image without consent, including artificially-generated images if a reasonable person would believe it actually depicts an identifiable individual. The second new law (AL HB 168) adds digitally created or altered visual depictions to child pornography laws. Both bills exempt from liability internet service providers, search engines, and cloud service providers if they simply provide access to the internet for such content.
Connecticut: The Senate passed a comprehensive AI bill (CT SB 2) and made major amendments (see above). The landmark bill is now before the House, which has less than two weeks to pass the bill and potentially make further amendments to get the governor on board.
Colorado: On Wednesday, the Senate sent two AI-related bills previously passed by the House back to that chamber after making amendments and passing the bills. The first bill (CO HB 1147) would prohibit distributing or publishing a political deepfake with reckless disregard that the candidate did not do or say what is being depicted within 60 days before a primary or 90 days before a general election without a disclaimer. The second bill (CO HB 1057) would prohibit the sale of algorithmic device services or products for the purpose of setting or recommending the amount of rent to be charged to a tenant.