Prince Charles (as he then was) once popularised the phrase ‘grey goo’ (Explained and Americanised by Wikipedia as ‘gray goo’). He wasn’t actually talking about AI but about nanobots. AI was merely a twinkle in a tech-bro’s eye at that point, but it sometimes it feels apt to describe the waves of samey AI junk we are all wading through, like so much verbal treacle.

A colleague sent me this interesting looking article on the increasing use of AI in US Courts by pro se litigants (that’s litigants in person or people without lawyers to UK lawyers and normal humans respectively):

Access to Justice in the Age of AI:Evidence from U.S. Federal Courts

I confess, because screens, the internet and social media have rotted my brain and destroyed my concentration span*, I haven’t read it in full, but it is on my ‘read it properly soon’ reading list, and it chimes with what I think we are all seeing on the ground in our own practices over here. The immediate response from colleagues who were cc’d was that they are definitely seeing a proliferation of obviously or apparently AI generated materials from litigants in person – and also sometimes from lawyers. This of course is not news, but the pace of change and the early signs of its real life impact on efficiency and justice is alarming.

Yes of course, AI is here to stay and it is sometimes (maybe often) helpful. But not always. And I am increasingly worried that when it is weilded by those without legal understanding, without an appreciation of its limitation of AI, and without an understanding of the expectations around AI use, or the risks and potential consequences of inappropriate use of AI or of reliance on its unhelpful output – it is not a force for good. I worry that rather than increasing access to justice it will create only an illusion of this, but in fact will generate worse outcomes for litigants and will generally increase the pressure and workload under which our justice system and our judges are already labouring. That’s the grey goo. Clogging up the system.

Statistics from Cafcass for the year to date (up to end April 26) show a stark and worrying trend over the last year. I’d be happy to bet a Solero** on that trend being, at least in part, derived from AI ‘assisted’ litigants. Of course, since we don’t really do proper statistics in the Family Court we’ll never know and I need never pay out on that bet, but it’s a reasonable hunch I think – at least consistent with the biggest uptick being in categories of case where there are the highest numbers of litigants in person (private law).

Interestingly, those stats are even more marked than the same picture for ytd up to end Mar 26 (no longer on Cafcass website but I happen to have downloaded this last month for a talk). I’m no statistician, but can’t help but notice that was was an 8.2% increase in the number of private law cases in March is now a stonking 15.9%.

What is driving that? What will it mean for case duration and waiting times? If a chunk of these new cases are driven by AI, will it also mean that those cases will be more….treacly? The emergent sense is very much that in many cases AI causes LiPs to produce verbose but ultimately unhelpful material, that consumes time and energy and distracts everyone from the task at hand to nobody’s advantage. It does not bode well.

Of course, it won’t all be AI driven, but I confess I’m not sure what else IS driving such a notable rate of increase of new apps. Could it be cost of living / economic issues? Or an increase in returners coming back to court after an order has failed? (if so, one hopes that does not include families returning to court after being through the Pathfinder process – it would be helpful to know one way or the other). Answers on a postcard please!

In more positive AI related news, this simplified judgment seems like a great use of AI:

Make it simple! Make it clear! Using AI to improve clarity and understanding of a judgment

I know from experience that when it comes to simplifying documents for those who need an easy read version, the aphorism ‘I didn’t have time to write a short one’ is always apt.

On a related note, I see that the BSB have issued guidance on use of AI by barristers, which is well worth a read. A consultation by the Civil Procedure Rules Cttee on the use of AI has just closed (although it didn’t cover use by LiPs), and the use of AI in Family Courts will most certainly be on the agenda at the Family Procedure Rule Committee in coming months (see March Mins), though whether such a committee can, with the best will in the world, be nimble enough to catch up or keep up with the pace of change is unclear.

Sliding neatly across from one set of Rule Committee Minutes to the next, and to the topic of QLRs appointed to ask questions on behalf of litigants in person who are prohibited from asking questions themselves, the April minutes of the Rules Committee make for alarming reading on a number of levels. The changes to the statutory guidance proposed, the implied and explicit reasons such changes are thought necessary and the problems under discussion in that meeting suggest that all is not well with the QLR system. Not well at all. Another source of ‘help’ for the litigant in person, which may sometimes not be much help at all, it seems. Particularly were it to come to pass that a QLR simply reads out the AI generated questions presented to them by the Litigant in Person, or uses their own AI generated questions…

*actually, I’m just trying to fit in some time sensitive focus-intensive work, some bank hol relaxation and a blog post today, so that is the corner I chose to cut.

** I have just been brought a solero by my lovely husband, and solero’s are on my mind, ok?