FAIR Act

This bill says federal agencies can only buy and use large language models (LLMs) that follow certain “trustworthy AI” rules. The models must be truthful when giving facts or analysis, value accuracy and science, and admit when the answer isn’t clear. They must stay neutral and nonpartisan. The text also says models should not be designed to steer answers toward any ideology, naming diversity, equity, and inclusion as an example. Developers shouldn’t build in political or ideological judgments unless a user asks for them or they’re clearly available to the user.

It sets a national goal to promote trustworthy AI and defines an LLM as a type of AI that makes natural-language answers from very large datasets. Federal agencies would follow these rules once the bill becomes law.

Key points:

• Who is affected: Federal agencies and AI companies that want to sell LLMs to the government.
• What changes: Agencies may only procure LLMs that are truthful, objective, acknowledge uncertainty, and remain neutral and nonpartisan; no built-in ideological steering unless prompted by the user.
• When: After the bill is enacted into law.