US Government's AI List Is a Mess—Here's Why It Matters

The US federal government publishes an AI inventory—a public list of AI systems it uses or plans to use. Ideally, this list helps the public see what AI is doing and who's responsible. But a new study looked at three years of these lists (2023–2025) and found that while the number of entries grew from 710 to 3,611, the information became less reliable. Each year, the questions asked changed, agencies classified things differently, and answers were overwritten—often without explanation. The result is what researchers call 'palimpsestic accountability': a record that's been written over so many times that the public is left to piece together the truth.

The study found that gains and losses in information were concentrated in a few agencies, and there was little documentation of human oversight—who checks the AI's decisions and how. This means that even though the list is public, it doesn't give a clear picture of how AI is being governed. For example, if an AI system denies your benefits or flags you for an audit, you might not be able to find out why or who to contact.

The researchers recommend practical fixes: versioned records (so you can see what changed and when), reasons for every reclassification, public files that distinguish between different kinds of missing answers, and oversight documented as conditions. These steps would make the inventory more useful for journalists, watchdogs, and ordinary citizens who want to understand how AI is used in government.

Why does this matter to you? Because government AI is already making decisions about benefits, immigration, and law enforcement. Without clear accountability, mistakes or biases can go unchecked. This study shows that transparency isn't just about publishing a list—it's about making that list understandable and consistent over time. As AI becomes more common in public services, keeping track of it is essential for fairness and trust.