Connecticut AI Legislation Marks a Starting Point, Not a Solution
Why It Matters
Connecticut has taken its first concrete steps toward regulating artificial intelligence, with two pieces of legislation addressing automated decision-making and AI-generated content. But policy analysts and legal scholars warn that the state’s new laws, while meaningful, represent a floor rather than a ceiling for protecting residents from rapidly evolving AI-related harms.
What Happened
Connecticut lawmakers passed two AI-related measures this legislative session. Senate Bill 5 was signed into law, establishing disclosure requirements and consumer protections around automated systems. House Bill 5312, placed on the consent calendar as of May 5, creates civil remedies for victims of nonconsensual AI-generated intimate images — commonly referred to as deepfakes.
Together, the bills reflect a broader recognition that artificial intelligence is no longer a future concern. It is already embedded in hiring decisions, workplace evaluations, consumer technology, and the flow of public information.
Senate Bill 5 targets some of the most consequential uses of AI in daily life. Its provisions include disclosure requirements when automated systems are used in employment screening, protections against algorithmic discrimination, notice rules for AI-powered companion applications, and labeling standards for synthetic media. The core principle is straightforward: people have a right to know when a machine is influencing decisions about their livelihood, their relationships, or their information environment.
House Bill 5312 addresses a more targeted but serious harm. AI tools have made it dramatically easier to fabricate realistic intimate images of real people without their knowledge or consent. The legislation empowers victims and the state attorney general to pursue civil action against bad actors and against platforms that fail to act when such content is reported.
By the Numbers
- 96% of deepfake content online consists of nonconsensual intimate depictions of women, according to a 2019 Deeptrace Labs report.
- 2 bills passed this session addressing AI transparency and synthetic media harm.
- May 5 — the date House Bill 5312 was placed on the consent calendar.
- The legislation covers multiple AI use cases: hiring systems, consumer subscriptions, AI companions, synthetic media, and economic planning tools.
Zoom Out
Connecticut’s approach mirrors a growing wave of state-level AI regulation across the country, as federal lawmakers have yet to pass comprehensive AI governance legislation. The European Union’s AI Act, which took effect in phases beginning in 2024, has set an international benchmark, while Canada has advanced its own AI and data regulation framework. Several U.S. states have moved on narrower issues — deepfake laws, algorithmic transparency in employment — filling a vacuum left by Washington.
Boston University Law Professor Danielle Keats Citron has described nonconsensual intimate image abuse as “terrifying, embarrassing, demeaning, and silencing,” adding that it signals to victims that their bodies are not their own. Her research has been influential in shaping civil-remedy frameworks that go beyond criminal penalties, which often arrive too late to stop the circulation of harmful content.
The deepfake problem is not limited to private individuals. Public officials and celebrities have also been targeted, raising broader concerns about AI-generated content distorting public perception during elections and other high-stakes moments.
What’s Next
Implementation of Senate Bill 5 will require state agencies and private employers to build compliance systems around its disclosure and anti-discrimination provisions. The attorney general’s office will play a central enforcement role under House Bill 5312, giving the state a public mechanism to pursue liability beyond individual victims acting alone.
Advocates and policy analysts argue that while Connecticut’s new laws establish a workable foundation, the pace of AI development means the legislature will need to revisit and expand these frameworks on a regular basis. Questions around generative AI in elections, automated systems in healthcare and housing, and the liability of AI developers themselves remain largely unaddressed at the state level.
Connecticut’s session has closed with meaningful progress. Whether the state continues to build on these measures — or treats them as a finished product — will determine how well its residents are protected as AI capabilities continue to advance.