Should courts ban AI evidence?

The Trailer

DavidPolice forces across England and Wales have paused using generative artificial intelligence to draft criminal statements. That pause followed an incident where West Midlands Police generated a ban describing stadium disorder that never happened. Civil liberties campaigners demand a total ban, while police technology leaders argue human checks protect essential efficiency. Should courts ban AI evidence? What's your opinion? Ban machine drafting in criminal trials, or allow verified tools to cut heavy backlogs?

The Facts

DavidPolice leaders across England and Wales have halted generative tools for preparing court statements. Meanwhile, courts and bar associations already punish lawyers who submit unverified, fake machine citations. Yet national police bodies are still building dedicated software to check case files. Tom, why should criminal courts ban these drafting tools completely?

Round 1 · The Case For

TomWest Midlands Police showed us the real danger when their software invented football disorder. Large language models work by guessing statistical patterns, not by checking real facts. Looking at the reality on the ground, asking a model to explain itself just creates another story. That means machine-drafted statements have no authentic trail of evidence. Some people think human review solves this problem. What makes this especially dangerous is that officers under crushing workloads trust smooth text. Checking every single sentence against primary files wipes out all time savings. If officers do not check thoroughly, hidden errors slip straight into court. When personal liberty is at stake, courts must ban generative evidence entirely.

DavidTom argues that statistical text generation creates fake details without a genuine audit trail. He also warns that full human checks remove any efficiency benefit. Emily, how do you answer that challenge?

Round 2 · The Case Against

EmilyPolice technology leaders like Alex Murray already insist on slowing down and verifying systems. Criminal courts already have strict authentication rules for all evidence. Lawyers and officers face formal professional sanctions if they present unverified claims. We do not need a blanket ban when accountability rules already exist. A total ban would also hurt practitioners who need help sorting through massive records. In civil medical negligence work, barrister Anthony Searle used these models to organise complex questioning. Separately, the firm Garfield AI used generated paperwork alongside a human barrister to recover unpaid fees. These cases show that software can assist case preparation when trained professionals oversee the work.

DavidEmily contends that existing courtroom sanctions and verification rules provide sufficient protection. She also highlights how legal practitioners already use tools to manage complex records. Tom, how do you respond to her argument on professional standards?

Round 3 · For Rebuts

TomProfessional ethical standards are a fair point, but rules alone do not stop errors early. That safeguard sounds reassuring, but procedural rules fail to catch mistakes before charging decisions. Look at what is happening in Derbyshire. A police detective is under criminal investigation for allegedly using chatbots to draft case paperwork. Several rape convictions are now under review because of those allegations. Waiting for courtroom appeals to catch flawed police drafting leaves innocent people in prison.

DavidTom points to the Derbyshire investigation to argue that ethical codes cannot prevent flawed files. He believes waiting for trial review comes far too late. Emily, what is your reply?

Round 4 · Against Rebuts

EmilyUsing unvetted consumer chatbots in police work is a genuine risk. That is why police forces paused informal uses while they build proper systems. The national program Police.AI is developing purpose-built tools to quality-check case files. They plan operational pilots in twenty twenty-six before any rollout in twenty twenty-seven. A total ban locks in crushing backlogs instead of building safe, tested tools.

The Crux

DavidToday's debate turns on whether human verification can catch errors without losing efficiency. Sanctions against lawyers show that professional rules have not stopped fake citations entering court. In six months, check whether the police drafting pause becomes a formal ban. That outcome will reveal whether courts trust procedural checks.

Your Turn

DavidNow, let us look at three useful debate expressions from today's discussion. First, Tom used this frame to enter an argument with practical context. Looking at the reality on the ground, officers face heavy daily caseloads. Looking at the reality on the ground, small legal teams need better tools. Looking at the reality on the ground, court delays hurt vulnerable victims. Next, Tom used this phrase to emphasise an urgent risk. What makes this especially dangerous is how easily false statements look real. What makes this especially dangerous is that unverified records cause wrongful arrests. What makes this especially dangerous is rushing changes without independent testing. Finally, Tom used this frame to concede before delivering a sharp reply. That safeguard sounds reassuring, but procedural rules fail to catch hidden mistakes. That safeguard sounds reassuring, but busy staff often skip detailed manual checks. That safeguard sounds reassuring, but statistical software lacks true accountability. Try using one of these three expressions in the comments below. Thank you for listening to The Breakfast Debate, and see you next time.