A recent study by a prison policy think-tank showed that there are an estimated 100,000 inmates for every one commutation of a sentence by a parole board.

Executive pardons by governors or even the president are even rarer.

While bureaucratic backlog — more than 160,000 federal incarcerated individuals currently under review — and a decades-long increase in incarcerations are often the blame, solutions have been few and far between other than simply hiring more parole board employees.

But developments in artificial intelligence may soon be able to help sort through the mountainous volume of cases and speed up the clemency process.

The Open Source AI Foundation recently launched the 20% Project, which is designed to use what’s known as “agentic AI” to review case documents and make recommendations to parole board employees.

Agentic AI acts independently of user oversight and makes decisions without human intervention — an electronic “agent,” as the name suggests. It can learn from its interactions and adapt to new situations, which is something that up until now it took a human mind to even begin to accomplish.

The hope, Open Source AI told The Hayride, is to save taxpayer money as well as bring justice to those waiting for clemency in jails and prisons across the U.S.

Led by Michael “Harry-O” Harris, founder of Death Row Records, who spent 33 years in prison before receiving a presidential pardon, the U.S. holds 5% of the world’s population but roughly 20% of its incarcerated individuals — “a statistic that demands urgent reform,” Harris said.

Now president of The 20% Project, Harris said he is hopeful lengthy, unnecessary prison stays like his will soon be a thing of the past.

“I was put up for pardon multiple times and denied due to emotional bias,” Harris said. “With open-source Agentic AI, we can overcome bureaucratic barriers while ensuring public transparency and participation. This technology allows us to work as fast as both humanly and ‘machinely’ possible, and I am honored to lead this effort for those still serving sentences.”

“Each pardon application requires an enormous amount of time and energy to properly address all of the questions and requirements, making the process slow and inefficient,” said Brad Cohen, general counsel for The 20% Project said, via a press release. “By leveraging open-source Agentic AI, we can dramatically accelerate this process while maintaining integrity and fairness. This initiative has the potential to right-size the criminal justice system.”

Brittany Kaiser, data rights activist and chairwoman of the Open Source AI Foundation, called the technology “the perfect partner for justice, transparency, and systemic change” in the release, while communications man James Davis hoped the initiative will “reshape the narrative around criminal justice reform.”

Agentic AI is in its early stages, according to a recent Reuters article concerning its emerging use in the legal field. The downside is that the technology presents “complexity for tech developers, who need to sit down with attorneys and sort through how exactly they come to a conclusion.” It also risks increasing the workload for lower-level employees who would have to sort through what recommendations and other data the agentic AI comes up with.

But the upside, according to the Reuters article, is an expedited legal system — though the technology is still young.

A 2003 analysis by University of California at Davis, of data obtained by the state of New York that, estimated that agentic AI “could have more than doubled the release rate without increasing the total or violent felony arrest rate.”