Caught using ChatGPT to write email

[Jidemummu199]

I don't normally use it to write emails just to structute blog outlines. So I had ever used it to write any text until then. I heard a lot of people have been using it for emails and I was curious to give it a try. Nothing else. The thing is what I didn't know the header and footer prompt responses were visible until I had already hit send about 15 minutes later. I immediately sent an brief apology to my colleague. We are normally very close and we think very highly of each other. So now I'm now concerned that this will make them think lesser of me. I will go into the office tomorrow and I'm nervous to see them. My husband tells me not to worry it happens. But do you think I'm worrying over nothing? One thing for sure is that I won't be trying it for emails again!

I'm not very hopeful that AI is going to be able to manage being a competent lawyer anytime soon. It's very far off right now and completely untrustworthy.

We aren’t allowed as we handle lots of personal data and the risk of a thwacking great fine for breach of gdpr is too high.

A year or two ago, all of the warnings about AI were that the machines were going to be smarter than us, and would take over despite our attempts to fend them off.

Now we can see that the real problem with AI is that we are too stupid to realise how shit the machines actually are, because we're enthralled by their approximations to expertise.

So we are blindly pushing ahead with outsourcing our decisions, not to super-intelligent versions of ourselves about whose benevolence we perhaps need to have questions, but to the IT equivalent of that bloke who always wins the pub quiz but never admits when he makes a really stupid error on one or two questions.

What are you on about?

I did use AI to generate the text. Out of interest, what seemed American about it?

Now this might well be a silly question, but if people stop going straight to a web search engine like Google, because AI is farming sources, how will those search engines continue, and where then will AI get its information from… just musings!
This is a really good point. When AI can consolidate content almost instantly, fewer people end up visiting websites or using google search, something advertisers will soon start to feel. At the moment, AI is refreshingly free of adverts, and I’d hate to see that change (even though it probably will).

As I was being made redundant anyway I didn't give a shit. Why would I waste my time writing emails and processes for the people who would be doing my job when I could use the company AI to do it for me?

Hi @MickyMoss it was “behavior” and to be honest just something about the way it was written but the spelling was the thing really.

I thought my question would be pooh poohed away but I was also thinking of advertising and that like many human hours it seems that search engines will be redundant… but not, and these are private organisations there to make profit.,

Interesting times but I’m glad I’m not starting out my working life as the AI age really takes hold.

Well yes I agree, but aren’t there parts of a lawyers job that could be more efficient with AI? There are dozens of people on this thread using it and so surely a lawyer isn’t so different that AI couldn’t help? Like I say, to give people better access to legal services at a lower overall cost? There’s nothing that law firms use right now? Really?

You need to use google to write an email?

Not just in the US. There was a case here just recently.

There was a case last month in the High Court where a barrister had used AI to come up with a list of fake cases - including one that purported to be a Court of Appeal case.

A local authority was opposing a homelessness claim. They lost the claim but were successful in getting a wasted costs order against the claimant's solicitors and barrister.

The barrister had come up with five fake cases.

There's an interesting article about it here:

https://www.legalfutures.co.uk/latest-news/judge-condemns-lawyers-who-produced-fake-citations-to-court

and a link to the judgment here

https://www.bailii.org/ew/cases/EWHC/Admin/2025/1040.html

Make sure tou have the privacy settings on, otherwise whatever you write is used as training data for the model for everyone else.
Also use the prompt to fact check everything it spits out with sources and citations.
Otherwise, I think you can spin it as you were testing it for productivity improvements.

No, not necessarily for e-mails, but we have all used Google for research, ideas, wording for communications etc etc.

We have moved on and so has technology. ChatGPT is just an advanced form of Google.

Well it’s still not a human so can’t replicate true creativity - which is what big companies pay the big buck for.

But for smaller marketing needs - promoting a local farmers’ market, or a one off small music festival, or your company’s annual charity event - no one is paying anyone £50K for that type of work so AI can maximize the time for that.

I am terrified of AI and no idea where to begin with it . Can anyone recommend a good training course for beginners?

On a tangential note, this shift in writing technology makes me wonder if people might return to the good old days of hand writing letters and postcards and then I remembered that Denmark's state-run postal service will cease letter deliveries by the end of 2025 as people simply do not post letters any longer for it to be a viable service. Somehow this made me sad, I hope the Uk won't follow this example.

I never use AI - can't stand it.

Some of my students have tried to slip it past me when submitting essays. I can always tell. It never occurs to them that their writing style is changed so much that it's glaringly obvious.

thebullshitmachines.com is a good starting point. It explains what current "Large Language Models" can and can't do, why they are sometimes useful but also why you should be very careful of relying on them.

@another2cats oh, so that has happened here as well now! I didn't know. (I did some training last year where they mentioned the US case, it seems that it's now cropping up elsewhere!). It was after that training that i started asking it questions about case law, precedents in certain areas etc...and got a mix of relevent real cases, irrelevant real cases and complete hallucinations. It makes for quite a feeling of panic when you realise that it presents these all equally confidently and makes them look real and relevant.

So I don't trust anything it says now. It's like a smooth overconfident bullshitter who likes the sound of his own voice and thinks if you say things articulately and arrogantly enough no-one will bother to check if they are all actually true! My least favourite type of person (I have come across a lot of them in my time!).

I won't use it for anything other than a casual chat or writing a haiku, until and unless it is much improved!

I always use chat gpt for my emails to check spelling , punctuation and tone. Heck Microsoft are already embeddeding ai proof reading and suggestion tools into their product lines. Don't sweat it.... Also you can tell chatgpt to remove the those header and footers.

Using AI for efficiency at work is generally considered a good thing but you must be careful with data and you have to edit it so it reads like a human wrote it. It’s really obvious if it’s purely AI.

There are already papers coming out in computer science showing that when AI is trained on AI-generated content, it goes rapidly downhill. Like a lot of technologies, AI is extractive: It takes current human knowledge and makes money off it, while leaving the externalities for our future selves to deal with. But as that cartoon of people huddling round a fire for warmth says, for one beautiful moment we created a lot of shareholder value.