The rules for AI are finally being written in Washington, and the stakes could not be higher.
For many years, there have been no rules about how the artificial intelligence (AI) business should work. When companies used algorithms to make hiring decisions, loan approvals, medical diagnoses, and criminal sentencing, there was no one federal agency telling them what was legal, what wasn't, or who was even watching. It looks like that time is ending.
The proposed Algorithmic Accountability Act, which was reintroduced in Congress with renewed bipartisan support in 2025, is the most ambitious plan yet to give the federal government control over AI systems that work in dangerous situations. If passed, it would be a big step forward for the economy, like what the Dodd‑Frank Act did for the markets after 2008.
What the Bill Would Do
The main idea behind the law is that companies bigger than a certain size would have to do impact assessments before putting in place automated decision systems that have big effects on customers. These evaluations need to look at risks related to fairness, privacy, security, and accuracy, and they need to be given to a specific federal body, most likely the Federal Trade Commission.
It is very clear in the bill what the difference is between general‑purpose AI tools and automated systems with big effects. A chatbot helping users draft emails would fall outside the framework. In this category would be an algorithm that decides who gets a job interview or a home loan.
Importantly, the bill would give the FTC the power to make rules about compliance, look into violations, and issue fines. This would fill a gap that has left regulators scrambling to apply consumer protection laws from decades ago to technology that the authors of those laws never thought of.
What It Leaves Unresolved
Like any sweeping legislation, the bill has limits. It doesn't create a comprehensive federal privacy law, so most of the rules about how to collect data come from different states. It does not directly regulate AI used by federal agencies or the military. It also avoids the more difficult question of who is responsible when an AI system hurts someone, leaving that to the courts.
Tech industry critics say that the need to follow rules could hurt smaller startups. Supporters of civil liberties, on the other hand, say that impact assessments that don't include real requirements for openness could just be checkbox exercises.
Why It Matters Now
The timing isn't a mistake. The U.S. has seen the AI Act of the European Union go into effect while American businesses are facing more and more lawsuits, hearings in Congress, and laws at the state level that go in different directions. Having a lot of different rules is now a risk for businesses, consumers, and America's ability to compete.
The Algorithmic Accountability Act won't resolve every tension in AI governance. But it would show for the first time that putting powerful automated systems into people's lives means the federal government is responsible for keeping an eye on them, and that responsibility has a permanent address in Washington, D.C.
