Connecticut Senate Bill 5 passed 32-4, imposing rules on frontier AI developers, creating a regulatory sandbox, and restricting AI chatbot use for minors.
Connecticut's Senate voted 32-4 Tuesday night to pass Senate Bill 5, a sweeping artificial intelligence regulation package heading to the House. The margin was lopsided; the debate was not. Legislators spent hours questioning whether the bill was too broad, too complex, and likely to drive investment out of the state.
SB 5, authored by Sen. James Maroney (D-Milford), would impose rules on "frontier" AI model developers, create a state AI sandbox for testing new products before full deployment, and restrict AI chatbot use for minors on social media. The bill also governs automated employment decisions and sets requirements for how state agencies use artificial intelligence.
"What we're doing is putting in important protections," Maroney said during floor debate. "Sometimes these machines get things wrong." Sen. Paul Cicarella (R-North Haven) crossed party lines to back the measure, saying he believed it "will do more good than any negative."
The hurdles ahead
The bill now faces the House and, beyond that, Gov. Ned Lamont. CT Mirror reported that a similar bill died last session under a gubernatorial veto threat, as Lamont argued that aggressive regulation risks Connecticut's economic competitiveness. His public position has not changed.
Connecticut has been circling AI legislation for years. Other state legislatures have drafted comparable bills drawing on the European Union's Artificial Intelligence Act, adopted in 2024, which introduced risk-based obligations for AI developers and gave American lawmakers a template. Few U.S. states have managed to pass anything substantial, citing competitiveness concerns nearly identical to those the governor has raised.
The industry is not waiting for legislators to catch up. CNBC reported in March that OpenAI plans to nearly double its workforce to 8,000 by year-end, with most new hires going into engineering, product development, and sales. TechCrunch noted this week that Google is expanding Gemini inside Chrome across seven Asia-Pacific markets. Artificial intelligence is being deployed in consumer products well ahead of any legal framework governing it.
Supporters of SB 5 argue that delay is no longer defensible. Forbes reported earlier this year that ChatGPT has reached 800 million monthly users and $20 billion in annual recurring revenue, figures that make political inaction increasingly hard to sustain. Maroney's coalition contends that residents deserve accountability from systems sorting their job applications, curating their children's content, and informing decisions inside state agencies.
The four dissenting senators were not persuaded. They argued the bill's scope was too ambitious and could deter investment, a concern the sandbox provision partly acknowledges: by carving out a protected testing space with relaxed rules, SB 5 implicitly concedes that some of its requirements may outpace what the market can currently absorb.
Whether the bill survives the House and reaches Lamont's desk in a form he will sign remains uncertain. Significant amendments are still possible.
A Connecticut veto would reinforce the argument that meaningful AI regulation at the state level requires federal action first. Passage would hand other state legislatures the most detailed domestic template in the country. The outcome matters well beyond Hartford.
---
FAQ
Q: What does Connecticut's Senate Bill 5 regulate?
A: SB 5 covers frontier AI model developers, creates a state sandbox for testing AI products under relaxed rules, restricts AI chatbot use for minors on social media, governs automated employment decisions, and sets requirements for state agency AI deployments.
Q: Will Connecticut's governor sign the AI bill?
A: Gov. Ned Lamont has not committed to signing it. He threatened to veto a similar bill last year over economic competitiveness concerns, and his position has not changed publicly heading into the House vote.
Q: How does Connecticut's approach compare to the EU Artificial Intelligence Act?
A: The EU AI Act classifies systems by risk level and imposes compliance obligations on developers, a model that influenced SB 5. Connecticut's bill is narrower in scope but follows the same logic of targeting high-capability frontier models first.
Q: What is an AI regulatory sandbox?
A: A sandbox lets companies test new AI products under relaxed regulatory requirements before full enforcement applies. Connecticut's bill would create one to support innovation while the broader legal framework matures.