Regulating AI Out of Nebraska
How AICPA tries to solve a problem that doesn't exist
This week I had the opportunity to visit Lincoln, Nebraska for the first time and testify on the Artificial Intelligence Compliance and Protection Act. We’ve been tracking AI bills across the country and pushing back on those most likely to stifle innovation, add unnecessary compliance burdens, or create vague, unworkable regulations.
Nebraska’s AICPA is, unfortunately, a prime example. While well-intentioned, its broad definitions and heavy-handed approach could drive AI development and deployment out of the state. Instead of fostering responsible AI use, it risks overregulating a technology that’s still evolving—and in the process, could harm Nebraska’s economy, businesses, and consumers.
There’s a better path forward. AI policy should focus on real harms, not hypothetical risks and ensure that regulation targets those closest to deployment, not the developers and open-source communities driving progress.
Hopefully my input to policymakers and stakeholders in Nebraska will help them arrive at a smarter, innovation-friendly approach. We are glad to work with policymakers and partners in the state of Nebraska and everywhere to get it right.
Below is a written version of the testimony I provided. I wrote the below version with Neil and when I was in Lincoln provided similar remarks.
Nebraska Judiciary Committee:
Thank you for the opportunity to comment. The Abundance Institute is a non-profit with offices in Salt Lake City and Washington, DC, focused on ensuring that lifechanging technologies—like artificial intelligence—reach their positive potential for human flourishing.
We are deeply concerned about the proposed Artificial Intelligence Compliance and Protection Act (AICPA). In its current form, the bill could drive businesses seeking to develop and use AI out of Nebraska. The U.S.’s longstanding permissionless approach to software development has made us the global technology leader. This bill would upend that approach, potentially depriving Nebraskans of significant health, education, and economic benefits.
We see two fundamental issues with the AICPA as drafted. First, the legislation is unnecessary. Federal and state laws already prohibit discrimination, regardless of whether AI is used. Imposing an additional AI-specific layer of regulation is redundant and risks undue burdens on businesses that are not engaged in discriminatory practices.
Second, the legislation is technically infeasible. Model developers, particularly those involved in open source projects, cannot predict or control applications and uses of their models. Moreover, the current draft defines artificial intelligence so broadly that it inadvertently includes basic computing and common software, forcing virtually every computing application into a new compliance regime.
The U.S. must lead in this technology, China is nipping at our heels, and Nebraska can help strengthen an innovation-positive regulatory environment. The Nebraska legislature should be laser-focused on facilitating the investment, development, and deployment of AI—a general-purpose technology that has widespread societal benefits. Yet, rather than addressing real, demonstrable harms through targeted regulation, the draft bill would impose a heavyweight, novel regulatory framework on a transformative technology based on hypothetical risks. We recommend that legislators address known harms, such as deepfake child sexual abuse material, through targeted legislation rather than attempting to mitigate all potential risks through a technology-specific regulatory approach.
However, if the legislature wishes to adopt a framework, three changes would significantly improve the AICPA:
Regulate the party closest to potential harm and benefit: the deployer who interacts directly with users. By focusing on deployers, the bill would better protect consumers and encourage innovation. Adding regulatory obligations on developers is unnecessary; any information a deployer needs can be addressed through contractual agreements between deployers and their suppliers.
Open model developers should not be burdened with obligations that are impossible to meet. The regulatory framework should clearly exempt open models—defined as those where developers make model weights widely available—from such requirements, ensuring that these vital tools continue to drive innovation and collaboration.
To prevent overbroad application of the bill, the definition of AI should focus on systems capable of autonomous learning and decision-making, excluding traditional, deterministic software.
Thank you for your time and consideration. We urge policymakers to adopt a balanced approach that safeguards Nebraskan citizens while promoting a thriving technological environment. We would be glad to work with the Nebraska legislature in these efforts.
Sincerely,
Taylor Barkley (Director of Policy) and Neil Chilson (Head of AI Policy)
 |
|