We Built an AI System to Find Unlawful Prison Sentences. It’s Already Getting People Out.
I build technology that works inside courts. My brother is a public defender. Together we built a system that finds sentencing errors and is already helping get people out of prison.
Casey should not have been there.
She was serving an 18-year sentence for a Class D felony. Under Alabama law, the maximum lawful sentence for that offense is ten years, with only two years required to be served day for day, meaning she would serve that time in full. When our system surfaced her case, she had already spent more than six years in prison and still had nearly twelve years left. This was not an edge case. She had a lawyer, the case moved through a major county, and a judge signed off on it. Multiple people reviewed it, and it still went through. That is the part that stuck with us. If something this obvious can make it all the way through the system, how often is it happening in cases no one ever looks at again?
This question wasn’t new for us. My brother, Adam Danneman, and I had been talking about it for years.
Adam is the Chief Public Defender for Jefferson County in Birmingham, where he leads one of the largest public defender offices in the country. Even running an office that size, he still digs into cases that don’t sit right. For years, he has been driving across the state on his own time and his own dime, reviewing cases in other counties and working to get people out whose sentences did not appear to align with the law. That is how he found Casey.
That case made something clear. If it could happen to her, it was likely happening elsewhere.
Last summer, we finally built something together, the first time we had actually collaborated. When we looked at this problem side by side, it wasn’t about legal expertise. Adam already had that. What was missing was visibility..
The data exists, but it is scattered across systems and nearly impossible to analyze in any structured way. Once someone is in prison, their case often fades into disconnected records. Appeals happen, but many cases are never revisited, even when something is off.
This is the kind of problem I’ve spent my career working on, taking fragmented systems and making them visible. So we built a way to see all of it.
We pulled together the Alabama Department of Corrections population into a structured dataset that included charges, sentences, and release dates. For the first time, it was possible to look across the entire system instead of one case at a time. We then used AI to surface what would otherwise stay buried, including sentences that exceeded statutory limits, structures that did not match the law, and cases where individuals had already served longer than required.
Casey surfaced through that system, and Adam took it from there. He filed the petition, worked the case, and got her out. She is now home, rebuilding her life with her daughter.
That result was not unique. We saw the same pattern repeat, including Sharon, a grandmother who had been serving a 30-year sentence for theft, and several others. Same pattern. Same outcome.
As we widened the lens, the bigger picture started to show up. We saw extreme outliers in sentencing and cases that did not line up with anything comparable. We saw patterns that appeared to vary depending on the county. We also began to see measurable differences across groups. In Alabama, early analysis suggests Hispanic individuals are receiving sentences roughly 30 percent longer than average for comparable offenses, and Black individuals about 9 percent longer. These are signals, not conclusions, but they raise questions that should not be ignored.
Another pattern was more straightforward. People who could not make bail and remained in custody were in a very different position than those who were released. When you are sitting in jail, the pressure to resolve your case is immediate. Waiting for a trial means more time in custody, so people plead. That dynamic alone can shape outcomes in ways that have little to do with the underlying facts of the case. At that point, this stopped looking like a handful of bad cases and started to look like something the system was not set up to catch.
Recent reporting, including the documentary The Alabama Solution, has shown what conditions inside these prisons look like. When someone is held longer than the law allows, every extra day matters. The problem is that there is no mechanism today that continuously checks sentencing outcomes at scale. Once a case is done, it is done, even if it is wrong.
We turned what we built into a system.
We call it Justice Audit. It is an AI system, part of my broader judicial technology platform 1Court, that analyzes prison data and flags cases that warrant legal review based on structure, statutory limits, or time served.
What began in Alabama is now running in Texas, Florida, Georgia, and Michigan, and is expanding quickly. The impact is straightforward. Many of the people we are identifying appear to have served more time than their sentences require, yet the system continues to spend public resources incarcerating them while they remain separated from their families and unable to work.
Adam approaches this through a justice lens. He sees the people and the families affected. I see that too, but I also see the inefficiency and cost. He leans toward the human side. I tend to think about how the system is misallocating resources, and what it costs when it gets it wrong.
There is a lot of talk about AI in legal right now. Some of it is real and some of it is not. The only thing that matters is whether it changes outcomes. This does. It does not replace lawyers. It makes it possible to find the cases that matter without relying on chance, and to do it across entire systems instead of one file at a time.
Since we began this work, leaders across the country have reached out. Court systems and advocacy groups are already exploring how to use this to reduce unnecessary costs and identify cases that warrant review.
Casey is out. Sharon is out. Others are out. More will follow.
The ability to audit sentencing outcomes at scale now exists. The question is whether we choose to use Justice Audit.
If you are working in courts, sentencing, or post-conviction review and want to be part of this work, reach out.
Noah Aron is the founder of 1Court, SimpleCase, and Sierra Alpaca Sanctuary.