When AI Gets It Wrong

efcbfca5

Hi All and welcome to Legal Compliance Academy's Pod Cast updates. The content of this podcast is for general information purposes "only" and "does not" constitute legal advice. With that disclaimer done, Buckle up everybody because: Here. We. Go.

efcbfca5

So, André, let’s get right into it. Regina versus The London Borough of Haringay—what a textbook case of everything going wrong when you rely on AI a bit too much?

c0eacccb

Absolutely. This case is a cautionary tale. Just imagine—a barrister presenting fake case precedents in court, and the explanation they gave? They called them, and I quote, "cosmetic errors". Honestly, it’s hard to believe.

efcbfca5

"Cosmetic errors!" Like, oh, I'm sorry, your brief has lipstick smudged all over it? What does that even mean?

c0eacccb

Exactly. The judge, Mr Justice Ritchie, was clearly unimpressed. He declared that providing five "entirely" fake cases amounted to professional misconduct. And while the underlying judicial review succeeded, costs were reduced by 7,000 pounds because of what can only be described as gross negligence by the legal team.

efcbfca5

Ouch. Seven grand down the drain because someone didn’t double-check their AI’s homework. I bet the firm wasn’t thrilled about that.

c0eacccb

No, not at all. And it didn’t end there. The judge ordered the hearing transcript to be sent to the Bar Standards Board and the Solicitors Regulation Authority. A pretty stern message to the entire legal profession. This case underscores why legal professionals "must" authenticate what AI produces. The risks of AI hallucinating, that is fabricating "plausible" but completely false information, are clear and obvious. Solicitors can’t afford to let their guard down.

efcbfca5

Or face the wrath of regulators, for that matter. Honestly, the more I hear about these incidents, the more I think, the rule: if it sounds too good—or bad—to be true, it probably is. So double, or even triple, check it.

c0eacccb

That’s good advice, Paul. With AI becoming more widespread, it’s all about balancing innovation with vigilance.

efcbfca5

And peaking of balancing innovation with vigilance, have you heard about Garfield Law? It’s being called the first fully AI-driven law firm in the UK. What's your take on that?

c0eacccb

It’s an exciting milestone. Garfield Law is trailblazing, but the SRA is watching them closely. They’re especially concerned about managing the risks of AI hallucinations, which they’ve identified as a high-risk area.

efcbfca5

Okay, wait, this term—“AI hallucination.” I'm an elderly gentleman now and I feel like I'm running to keep up! What on earth is an "AI Hallucination"!

c0eacccb

You're not that old, just a bit decrepit! So Paul, an AI hallucination is "essentially" when the system generates information that "sounds" plausible but is "entirely" false. And when it comes to legal practice, that’s a disaster waiting to happen.

efcbfca5

That sounds really "dangerous", possibly even reckless?

c0eacccb

You're right. Even the SRA has forbidden Garfield Law’s AI from proposing relevant case law to avoid this issue. This is critical because legal professionals rely heavily on accurate precedents and citations. AI hallucinations undermine that trust entirely. And remember, Paul, the SRA Code of Conduct states that you have a "professional obligation to uphold the reputation of the profession and maintain public trust"

efcbfca5

Really good point, André. Did I read somewhere that the SRA reported three-quarters of the largest solicitor firms are already using AI? Double what it was just three years ago?

c0eacccb

Yes, that's true. But usually in a supporting role because the stakes are especially high, given the risks of professional misconduct if they fail to verify AI-generated outputs.

efcbfca5

So if you were working in a solicitors firm would "you" trust AI?

c0eacccb

Unless I had verified every single output? Not a snowballs chance in hell! AI is a tool, but it doesn’t replace critical human judgment. Innovation is essential, but it needs to be paired with robust checks and balances.

efcbfca5

So why "are" big firms starting to use AI tools?

c0eacccb

Well AI offers efficiency, speed, and cost-saving potential that’s hard to ignore. But as with any powerful tool, there’s always a flip side—especially when the stakes involve professional misconduct.

efcbfca5

So what should firms be doing?

c0eacccb

Firms have got to adapt—work out if and where AI might help their practice, introduce "specific" policies to address AI-driven risks, and give guidance to staff.

efcbfca5

Wait, policies? Like what?

c0eacccb

Well, it's about embedding traditional verification measures into workflows—checking AI-generated citations with the "same" rigor used before AI was in the picture. It’s a way to balance innovation with good old-fashioned Yorkshire diligence, Paul.

efcbfca5

Ahhhhh that “good old-fashioned Yorkshire diligence” you're always telling me about!

c0eacccb

Yes! Exactly! I've already seen Professionals learning to adapt, and regulators like the SRA are stepping in to help guide the process. It’s all about finding that balance between embracing the benefits of AI, and maintaining accountability.

efcbfca5

Well, here’s hoping they get it right. In the meantime, maybe we should leave ChatGPT out of our legal services, rename it to ChatNot-QuiteGPT. What do you think?

c0eacccb

I think you're kids are right about your Dad jokes. But it's a good reminder to double or triple check everything it says. Just don't blindly trust it.

efcbfca5

Like you never trust anything I say! So that's it for this time, everyone. Thanks for tuning in. Remember, if life gives you AI hallucinations, do what André says—verify, verify, verify.

efcbfca5

Until next time, take care, and stay legally compliant!