Regulatory restrictions on evolving technology may provide a sense of safety from potential harm, but they also stymie modernization. A novel creation just adopted seven years ago in the United States, a regulatory sandbox temporarily suspends some of these legal barriers, while still maintaining government oversight.
This method of supervision enables entrepreneurs to safely innovate while getting consumer feedback, without continually submitting applications to regulatory bodies for every new product development. One emerging field especially suited to this structure is artificial intelligence (AI).
Unique AI Needs
According to federal law, AI includes “any artificial system that performs tasks under varying and unpredictable circumstances without significant human oversight, or that can learn from experience and improve performance when exposed to data sets.”
Several potential dangers make themselves immediately apparent in that brief explanation. Such an apparatus performs actions on its own, but can never truly develop the moral judgment humans must use in choosing behavior. It learns from experience, but its experience can never duplicate human life. It requires data sets to improve, and data often includes sensitive information. It can access this private data without permission.
On the other hand, the technology that enables AI also provides enhanced methods of protecting privacy, and the potential speed of innovation that makes it so effective slows if restrained by outdated paperwork processes.
In taking these factors into consideration, states and the federal government are using sandboxes to develop a variety of ways to create appropriate boundaries.
Disclosure
Even people who welcome AI assistance might want to know if they are interacting with a person or a machine. The Utah AI Policy Act (UAIPA) was passed in 2024, and by the next year, there were already amendments containing enhanced rules ensuring this disclosure.
The current Utah law specifically mandates that consumers have access to this knowledge if they request it from a company using GenAI, defined as a program that “is trained on data, interacts with a person in Utah, and generates outputs similar to outputs created by a human.” If the company involves health care or other safety-sensitive issues, it must reveal the AI use even without a customer query.
Restrictions On Intent
The Texas Responsible Artificial Intelligence Governance Act (TRAIGA), which goes into effect in 2026, aims to prohibit using AI for illegal intent such as unlawful discrimination, child pornography, and harmful deepfakes.
Working in a sandbox enables participating companies to eliminate algorithm flaws that might inadvertently lead to these outcomes. Instead of facing private lawsuits, they have 60 days to reach compliance and settle the issue with the Texas attorney general.
Federal Oversight
At the federal level, Sen. Ted Cruz recently introduced the Strengthening Artificial Intelligence Normalization and Diffusion By Oversight and Experimentation (SANDBOX) Act. If passed, this legislation will prevent federal laws from overriding helpful state-level sandbox laws.
The SANDBOX Act also provides a modern framework to coordinate decisions across multiple federal agencies, facilitating smooth flow of paperwork to appropriate regulators. The Office of Science and Technology Policy (OSTP) will oversee the sandbox and refer regulation waiver requests to the relevant federal body, which will return a decision within 90 days.
Facing The Future Without Fear
Regulatory sandboxes and AI are both new and rapidly evolving, and the combination of the two in the United States has only existed for a year. Legislators must take seriously their constituents’ concerns about the potential consequences of burgeoning technology, while also emphasizing the opportunity cost of restriction.
Ironically, the overzealous insistence on security can easily prevent innovations that would actually increase safety for consumers. Lawmakers have the duty to ensure this does not happen with AI, and a sandbox provides a way to prevent it.