In partnership with
You can be compliant by Friday with one line of text. Two things happened this month that most people using AI for work haven't caught up with.
On the 2nd, the EU's transparency rules became enforceable. If a person is talking to your AI, you have to tell them.
On the 14th, Anthropic started watermarking Claude's text output. Invisible, but detectable.
Put those together and the quiet plan a lot of people were running on, which was that nobody would ever know, stopped being a plan.
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.
Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:
7 high-impact AI strategies to accelerate your marketing performance
Practical use cases for content creation, lead gen, and personalization
Expert insights into how top marketers are using AI today
A framework to evaluate and implement AI tools efficiently
Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.
The disclosure rule in three lines
It's Article 50 of the EU AI Act, live since 2 August 2026. The parts that touch a small operation:
Chatbots. If people interact with an AI system, they must be clearly told, unless it's blindingly obvious.
Synthetic content. AI-generated or AI-modified images, audio and video need machine-readable marking.
Deepfakes. If it depicts real people or events and it's fabricated, it must be disclosed.
Two details catch people out.
First, it applies based on who you're talking to, not where you're registered. If your bot messages anyone in the EU, you're in scope from a kitchen table in Ohio.
Second, using somebody else's chatbot doesn't move the obligation. Drop a third-party widget on your site and you're the deployer. It's your notice to write.
Fines run to €15 million or 3% of worldwide turnover, whichever is higher. For small companies the lower figure applies, which is the one concession in the whole thing.
What Anthropic's watermark can and can't tell
Anthropic's announcement on the 14th is the other side of this.
Claude's text now carries a statistical watermark, built on the same approach as Google DeepMind's SynthID. You can't see it. It survives light editing. Anthropic says a detection tool is coming.
Worth knowing what it doesn't do. It doesn't identify you, your company, or which conversation the text came from. It doesn't get applied to things with one right answer, like code or a calculation. Files get a content credential tag instead.
So it's provenance, not surveillance. But it does mean "AI-assisted" becomes a checkable claim rather than an honour system, and that's a real change for anyone selling writing.
What to actually do this week
Three things, and none take long.
If you run a chatbot or voice agent, put a plain line at the start of the conversation. "Hi, I'm an AI assistant for [business]. I can help with X, or put you through to a person." That's the whole obligation.
If you publish AI-generated images or video, keep the labelling the tool applies and don't strip the metadata.
If you write for clients with AI in the loop, decide your disclosure line now and put it in your terms. Being asked and having an answer ready beats being asked and improvising.
Here's a prompt to work out where you actually stand:
You're advising me on AI transparency obligations. I want a practical checklist, not a legal essay.
My business: [what you do]
Where my customers are: [countries or regions]
Where AI touches my work:
- [e.g. a chatbot on the site]
- [e.g. AI-assisted copy for clients]
- [e.g. AI-generated images in marketing]
For each one, tell me:
1. Whether a disclosure obligation plausibly applies, and under which rule.
2. The exact wording I could use, written for a normal person.
3. Where it needs to appear.
4. What I should keep as a record.
Flag anything that depends on specifics you don't have. Say clearly where I'd need a lawyer rather than guessing.I'm not a lawyer and this isn't legal advice. It's a starting map, and for anything with real exposure you want a professional looking at your specifics.
Find out before a customer does
Rules in this area are going to keep moving for the next few years.
A monthly scheduled task that checks for changes to AI disclosure requirements in the markets you sell into costs you nothing and means you find out from your own briefing rather than from a customer complaint.
You can sort all of that this afternoon.
The bigger question is what you charge for knowing it, because plenty of small businesses have a chatbot on their site and no idea any of this happened.
That's the second half, and it's the whole reason the community exists.
Quick one.
Do you disclose when AI helped? Hit reply, I'm curious where people land.
Talk soon,
Zephyr
