Conn. sets foundational standard for AI regulation

Governments worldwide are already taking action on AI — including in the European Union with the passage of its landmark EU AI Act — policymakers at all levels are anxious and eager to pass legislation of their own to protect citizens and businesses from the most harmful uses of AI, such as misinformation.

While Congress sorts out the next steps on the federal level, individual states are rightly forging ahead with their own versions of AI regulation. Connecticut’s most recent bill, Senate Bill 2 (SB2), stands out as a beacon of foresight and balanced regulation. At its core, this bill understands that not all AI applications are created equal. This is critical for three reasons.

First, this approach acknowledges the vast spectrum of AI capabilities, from benign applications in customer service to more potent uses in health care, finance, and law enforcement. What makes SB2 so relevant is its detailed approach to regulation, which mandates stricter scrutiny and oversight for AI systems that pose higher-level risks to public safety, privacy, and society as a whole while still allowing for innovation and growth in areas where its applications are lower-risk. The bill requires developers of high-risk AI systems to conduct impact assessments and disclose potential risks, including biases, errors, and the extent of human oversight. This transparency builds public trust and encourages developers to preemptively address potential issues, leading to safer and more ethical AI systems.

Second, this framework ensures that innovation is supported by avoiding broad regulations that fail to distinguish between high-risk and low-risk AI applications. In the tech industry, where innovation is rapid, blanket regulations can quickly become obsolete or unnecessarily burdensome, hindering growth and competitiveness. SB2’s targeted regulatory framework ensures that innovation is supported by allowing AI developers to innovate freely in areas with low risk, while simultaneously protecting consumers.

And finally, focusing on high-risk applications ensures that regulatory resources are allocated efficiently, concentrating oversight where it is most needed to protect the public. By identifying and scrutinizing AI systems with the potential to significantly impact society — such as those affecting critical infrastructure, personal data, or having the ability to influence legal outcomes — SB2 prioritizes public safety and ethical considerations without overextending regulatory resources. For example, this bill requires employers to inform workers of any automated decision-making tool being used to make consequential decisions, as well as explain its purpose and provide a clear description of the tool.

We need this type of precision regulation if we’re going to strike the right balance between protecting society from potential harm and allowing innovation to flourish. We must ensure that regulatory approaches get to the heart of the issue and won’t stifle society’s ability to realize the full benefits of AI.

Connecticut’s risk-based framework should serve as an example of a positive path forward for AI regulation at the state level, and other states across the country would do well to follow suit. Not all regulation will be — or needs to be — identical. Still, similar common-sense principles will yield AI regulations that are effective and interoperable between jurisdictions, which is important across state lines in the United States.

Additionally, the private sector must work quickly and proactively to put in place proper governance standards and practices for technologies such as AI. It’s the role of industry partners to work with the public sector to ensure businesses are held accountable and society is able to reap the benefits of AI while addressing the potential for misuse and risk.

This is a pivotal time in AI, and every state should be looking at Connecticut’s foundational standard and example for AI regulation. As we stand at the cusp of an AI-driven era, the need for thoughtful, flexible, forward-looking regulation has never been greater.