Regulate new AI programs that make critical decisions by requiring they first pass a government test
Currently there is no federal requirement that AI programs used by government agencies or in critical areas be tested before deployment. A proposal would require that AI programs be tested and approved before they are used, similar to how the government requires drugs to be tested before they are approved.
Arguments For & Against
Pro Argument
It is the government's duty to prevent harm. Just as the government requires testing of new drugs before they can be sold, it should require testing of AI programs before they are deployed. Companies cut corners to get products to market faster, and profit motives mean safety can be deprioritized. AI systems have already wrongfully accused thousands of people of crimes.
Con Argument
The government should only intervene after there is evidence a specific AI program is causing problems, not before. Requiring pre-approval will slow innovation and give other countries a competitive advantage.
Source document: AI_Quaire_0224-1.pdf
| Type | Organization | Date | Nat | Rep | Dem | Gap | Metric |
|---|---|---|---|---|---|---|---|
| New PPC Survey (2026) | Program for Public Consultation | February 2026 | 78% | 78% | 82% | 4% | favor |
| Deliberative Survey | Program for Public Consultation | February 2024 | 81% | 76% | 88% | 12% | favor |
Program for Public Consultation — February 2026
The first set of proposals addresses AI programs that make decisions that can have significant impacts on people (e.g. in health insurance claims, loan applications, hiring, etc.) To ensure these AI programs follow regulations, have data privacy and security protections, and do not have unintended biases, do you favor or oppose: For new AI programs, require they first pass a government-designed test.
Program for Public Consultation — February 2024
Require pre-testing for any new AI program that is going to be used to make decisions that can have significant impacts on people, including in healthcare, banking, housing, education, employment, legal services, and utilities (like electricity). These tests would try to ensure that the AI program: follows regulations and best practices, has data privacy and security protections, and does not have unintended biases. These tests would be conducted by the government or an independent third-party. If it does not pass the tests, it would not be approved for general use.
Related Policies
Create a new federal agency to monitor new AI programs, anticipate potential problems, define best practices, and enforce regulations
80% national support
Allow the government to audit, on an ongoing basis, AI programs that make critical decisions
74% national support
Require all AI-generated deepfakes to be labelled as such
86% national support
Prohibit political campaigns, including PACs, from using deepfakes
82% national support