Everyone in the world is thinking about AI right now, though not all in the same capacity. Some people are embracing it head on, fully integrating AI tools into their daily life at home and at work. Some are terrified of it – losing jobs, losing privacy, robots taking over. Some are in the ideological and ethical debate, others are in the minutia of its application in business, government, and other parts of society; and some just want a good chatbot, and I don’t blame them! Most people, however, probably don’t spend all day thinking about how AI might impact the entire criminal justice system in America. (If you do, let’s talk!)
As a longtime justice reform advocate and philanthropist, the question for me isn’t whether AI will be infused into justice (it already is), it’s how we embrace it in this sector that matters more than anything. The way justice and AI collide today will impact the way our entire society functions in the future.
In the spirit of the upcoming 250th birthday of the United States in July, I asked Claude a question about the future: what will the criminal justice system be like 250 years from now if we do nothing. Specifically, if we don’t disrupt current models of criminal justice in America, and if we also continue the current development and investment trajectory of justice and public safety AI tools.
Here is Claude’s prophecy:
The criminal justice system of 2275 will be almost unrecognizable in its mechanics. Near-total surveillance and behavioral AI will make unsolved crime rare. Algorithmic systems will handle most prosecutions and sentencing. Physical prisons will largely disappear, replaced by digital confinement and automated restitution. Rehabilitation will become the system's primary stated purpose, backed by neurotechnology capable of genuinely altering the impulse patterns and emotional architecture that drive people to offend.
If this isn’t something to sit with, I don’t know what is.
I call this response a prophecy not because chatbots are oracles, but because I’ve seen versions of this premonition all over the justice space, and we’re already living into it. Claude has just been listening. As large language models do, it’s reflecting and projecting the vast library of knowledge that humans have put online. The hopes, fears, histories, and importantly, capabilities, of people and nations for generations.
AI itself is showing us one path to a massive reduction in physical jails, prisons, police, the court system, and crime, should we stay the course. At the same time, it’s also showing us what we may have to give up to get there: freedom, agency, physical autonomy, and personal privacy. Many people might be intrigued by this concept at face value, and move quickly to make it real (because problem solved, right?) Others may run for the hills to evade the potential surveillance state. I land somewhere in the middle – probably like many Americans. Terrified and excited. Cautious but optimistic. Deeply craving transformation to an archaic system, alongside some healthy skepticism and a set of checks and balances, clear guardrails, and a commitment not just to the destination, but to the process.
AI certainly can – and will – expand the grip of the justice system on American lives if we let it. Crimes are surely solved faster when everyone is a suspect. But AI could also become the great equalizer of American justice – helping the U.S. achieve the most transparent, accountable, equitable, and laser-focused justice system the world has ever seen. A new model to emulate on a global scale.
We get to decide. People who care about, work in, are responsible for, and are impacted by the justice system still have agency in shaping its future.
The right tools, deployed in the right ways with the right guardrails…this will be the difference between the “eyes everywhere” crime reduction strategy described by Claude, and one where the justice system is a precision tool, a last resort intervention, and a powerful stepping stone for people to heal and become accountable – should anyone need to meet it on their worst day.
The good news is there are already meaningful examples of human-centered justice AI out in the world. They just need investment, replication, and to be plugged into a well thought out tech stack across the justice system. They also need AI’s best and brightest working on them.
There are dozens of new tools that analyze prosecutorial charging patterns, and audit them for disparities, irregularities, and biases. Products like JusticeText are streamlining legal work, helping public defenders do things like quickly transcribe bodycam footage and identify inconsistencies in police statements for fairer representation. The Center for Employment Opportunities has an AI tool that coaches incarcerated people in preparation for job hunts once they’re home. Recidiviz’s Case Planning Assistant helps correctional departments develop personalized plans for people in custody – mapping in-facility programming to post-release housing and transportation, and giving people access to their own records.
There are also groups working to establish principles and guardrails for justice AI. The Council on Criminal Justice’s Task Force on Artificial Intelligence has developed “standards and evidence-based recommendations to guide the safe, ethical, and effective use of AI in the criminal justice system,” such as ensuring there is always human oversight. The AI and Justice Consortium brings practitioners together to share knowledge and improve the implementation of new AI tools.
Imagine all of these tools (and many more), working in coordination across every part of the justice system against strict guardrails, driving better, measurable outcomes for people and communities every step of the way. That’s systems change right there.
This kind of intentional, hands on integration gets us closer to a world where more crimes are solved and deterred without widening the net into the justice system. Where people who have committed very real crimes are held accountable, and then have the opportunity to move through a carefully mapped, transparent, efficient journey of healing and transformation, should they choose to engage. Where sentences are driven by data not vengeance, and are proportionate to the crime committed. Where access to powerful tools of rehabilitation and skill building are democratized and readily available to all. Where fewer people endlessly cycle in and out of jail, prison, and a life of crime (because AI is revolutionizing rehabilitation, healing, and reentry readiness). Where people feel safe without feeling surveilled.
Of course, getting to this place will take a ton of work. And, I already see us leaning into Claude’s prophecy with curiosity, and a lot of resources. Experimental dabbling in AI tools quickly slips into massive contracts. Police departments starting with largely helpful body cams and license plate readers, now investing tens of millions in widespread surveillance tools (which, according to Claude, will eventually put those same officers out of a job). Predictive models being used in policing and sentencing; machines intending to remove human bias but often deepening it, while also removing human empathy, compassion, context, and reason.
Advocates like myself have been pushing for a massive overhaul of the justice system for a long time, but we aren’t asking to replace it with expansive new forms of state control. We’ve demanded real solutions to crime and safety, but not at the expense of Americans’ personal freedoms and privacy. We’ve pushed for better rehabilitation and accountability, not medical procedures to engineer concerning behaviour out of our kid’s brains.
The most important first step is not to accept anything as inevitable. Instead, we must treat Claude’s prophecy as a warning; as a choice about the world we want to live in. AI for the sake of AI and profit gets us the surveillance state the bot describes. Principled AI centered on humanity, equity, dignity, and privacy gets us closer to justice than we’ve ever been.
Navigating this fork in the road with care is my commitment for the next 250. To use every tool in my toolbox to ensure AI enters the justice system in the most humane, transparent, accountable, and data-informed ways possible – for the benefit of all Americans. To shrink and improve, not expand and deepen.
