# What This Bill Does

The federal government wants to create special testing labs to make sure artificial intelligence systems work safely and fairly before government agencies use them. Right now, when federal agencies adopt AI tools—whether for processing benefits applications, managing databases, or making decisions about services—there's no standardized way to check if these systems are trustworthy or if they might discriminate against certain groups of people. This bill would fix that by establishing testing centers run by the Department of Homeland Security to evaluate AI systems before they're deployed across government.

The testing labs would specifically check whether AI systems protect important things like privacy, fairness, and people's right to human decision-making. For example, if an agency wants to use AI to help decide whether someone qualifies for benefits, the labs would test whether the system treats everyone fairly and doesn't automatically deny benefits without human review. The bill also requires that AI systems can't make final decisions about denying government services on their own—a human has to be involved. Additionally, the bill creates a digital library of real-world test cases (without anyone's personal information) that agencies can use to train and test their AI systems in realistic scenarios.

The bill also establishes a new office within Homeland Security where federal agencies can report problems they've experienced with AI systems they're already using. This creates a way for agencies to learn from each other's mistakes and share solutions. The government would need to report back to Congress every six months about how the testing program is working, what challenges they're facing, and what resources they need. To make all this happen, the bill authorizes $20 billion in funding.

To provide for Federal civilian agency laboratory development for testing and certification of artificial intelligence for civilian agency use, and for other purposes.