This was a Talking Web newsletter on 18 August 2026 - join the newsletter here.
Hi,
Happy Tuesday! I hope you’re having a lovely August. We’re just back from a week in Britanny which was lovely – even though I came back to over 1100 emails 🫣. I’m currently spamming my personal Insta with photos if you want a sneaky peak.
Many of those emails were about the EU’s Article 50 – have you heard of it?
It’s the ruling that came into force on the 2nd August, stating that certain sorts of AI content needs to be clearly declared as being AI.
Now, please insert your own choice of legal disclaimer here(!) 😜… I’m just Lisa at 18a making websites, you should get your own legal team to give you actual concrete advice on how you adhere to Article 50, but here’s what I’ve read so far:
(Do give #2 careful consideration).
(And remember that this *probably* only concerns you if your customers are in the EU.)
#1 Text
If you’re writing content about public interest matters, such as politics, health care or other “important” stuff, and you’re just getting Claude or Chatty or another to churn it out without reading it and it’s being set live automatically – that needs to be marked as being written by AI.
If you’re writing marketing material, sales emails, case studies for your website or opinion pieces about your not-important-to-the-EU industry, then you’re OK. Especially if you’re reading it, editing it before publishing it, and putting your name to it.
It seems to be certain topics of content, being automatically written and published without human input that the rules are aimed at.
Which I think is a pretty narrow remit really.
#2 Images
Now this one might catch more people…
If you’re making any sort of deep fake images, or AI video or audio, then they need to be flagged. And the EU even made little icons you put on them.
But what does “deep fake” mean?
And that’s the bit that’s got some wiggle room for interpretation…
This good read defines a deep fake in this context as an “AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful.”
BUT then goes on to say… “Even an abstract resemblance to fictional persons or objects may suffice, provided the content could be perceived as authentic or truthful. “
“may suffice”… that’s where things start to get murky.
SO – does that then include made up photos of unreal people sitting in your restaurant, or enjoying your training course?
You’re not creating a deep fake of the Prime Minister or a pop star, or even yourself (cos people using AI to make avatars of themselves for speaking videos is definately included) – you’re creating people who don’t even exist.
This means your job for this week is to consider if you’ve got any photos like this on your website or in your marketing materials, and if you have, give them a little icon or identify them in some clear way.
#3 Chatbots
Also just a note that if you’ve got some sort of automated customer service which uses AI, this needs to be declared now too.
Someone I follow on Insta has, for a long time, sent her ManyChat messages from “Lisa Bot, Lisa’s AI assistant” (but with her name, obvs). And I think that’s a great idea! It’s always felt completely transparent, which is a refreshing change.
BUT – do we think that the publications who just automatically churn out political stuff or fake pics of politicans to grab headlines and stir up debates will worry about these rules? I’m not convinced…
I actually liked a lot of the GDPR, and I REALLY think AI needs governanace… but I’m not convinced these rules are going to really help anyone, and will just cause more of a headache for the people who take these sorts of rules seriously (who probably aren’t the people using it in this way anyway).
Please do let me know your thoughts – hit me a reply, and let me know if you’ve needed to update anything on your website.
Speak soon,
Lisa
Lisa Freeman