Transparency

oriiion and the EU AI Act

There is a new EU law about AI. Here is what it means for you: what our tool does, what you decide, and what you can switch off.

Last updated: 2026-08-19

Start here: the questions we actually get

Each one links to a fuller explanation further down.

Can I use AI images in my marketing?

Yes. There is no ban on AI images in advertising. The rule is simply: do not fool anyone. If a picture looks like a normal photo, and shows people, places or events that could be mistaken for real, add a line saying it was made with AI. If it is drawn, illustrated or clearly not a photo, you do not need to do anything. And if you would rather not think about it at all, you can switch AI images off and use only your own photos.

Can I keep using oriiion?

Yes. The strict parts of the law are about things like AI that screens job applicants or decides on loans. Posting on social media is not in that group. Two things need to be visible rather than hidden. First, that our assistant is a robot and not a person. We handle that. Second, that a realistic AI picture is flagged as AI. That one is your call when you post.

Do I have to label every text AI helped write?

Almost never. The labelling rule for text is aimed at news and public-interest reporting. Ordinary marketing and advertising is not covered. On top of that, the rule falls away when a person reads the text and approves it before it goes out, which is exactly how oriiion works. So it is the pictures to keep an eye on, not the words.

Can oriiion post something without me seeing it first?

No. Everything we make is a draft. It only reaches your accounts once you press approve. A scheduled post is one you already approved for later. The only thing that sends on its own is an auto-reply you set up yourself, and there the wording is text you wrote.

Can I switch AI images off completely?

Yes. One setting turns off image generation for your whole account. With it off, nothing is made up. Your posts use your own product photos, pictures you have uploaded, or stock photos. Just ask us and we will switch it on for you.

Can I switch off the AI writing too?

No. Writing captions and post ideas is what oriiion is for, so there is no off switch for that. If you want no AI involved in your marketing at all, oriiion is not the right tool for you. What you get here is: AI writes the first draft, you decide what goes out.

Are your AI images tagged so Instagram and TikTok spot them?

Do not count on it. Many AI tools add an invisible tag inside the picture file. That tag is wiped when a picture is cropped or redesigned, which happens often in our tool. Some tags sit in the picture itself and survive better, but we have not tested that and will not promise it. Assume the platform will not flag it for you, and write it in the caption yourself when it matters.

If something goes wrong, is it my problem or yours?

Depends which part. If our assistant failed to say it was a robot, that is on us. If you posted a realistic AI picture of a real place without saying so, that is on you, because you are the one who published it. No supplier can take that part off your hands, and you should be wary of any that says it can.

Who sees my content, and do you train AI on it?

Four companies, and no. Your content goes to the AI companies we use, purely to answer your request. We do not use it to train anything, and we build no AI of our own. Every company we use is named on our suppliers page, along with what it does and which country it is in.

Where is my data kept, and can I get it back?

Yes, and we say where. You can download all your data and delete your account from the privacy section of your settings. Where the database physically sits is written in our privacy policy rather than hidden.

What about images I upload myself?

Those are yours. We do not inspect what you upload, and we cannot tell whether a file you supply was made by AI somewhere else. So an uploaded image is your responsibility: if it is AI-made and could pass for a real photo, it is on you to say so. We always treat an uploaded file as a real photograph. If one of yours was made with AI, you can mark it in your gallery, where you can see the picture you are describing, and we pass that on to the platforms that accept the flag.

The law, and when it started

The EU's AI law started applying to this kind of thing on 2 August 2026. Our AI transparency notice says the same things in formal legal wording.

1. What we do to keep you safe

This is what the tool does today.

Nothing goes out on its own. Every post is a draft until you approve it. A scheduled post is one you already approved for later. The AI never decides to publish.
Our assistant tells people it is a robot. That goes for the app, the website, Telegram, WhatsApp, Messenger, SMS and the phone line. If you would rather deal with a person, email our support address below.
We do not train AI on your content. It goes to the AI companies we use to answer your request, and nothing else.
We tell you what our image files really contain, including when the answer is annoying.
The switch for AI images works everywhere posts get made, not just on one screen.
We publish the full list of companies we use, and update it in the same go as adding one, so the page cannot quietly fall out of date.

None of this means the AI is always right. It writes confidently and is sometimes wrong, especially about prices, dates, opening hours and the details of your own products. That is why you check before it goes out.

2. What is inside the image files

Many AI tools put an invisible tag inside a picture file to show it was made by AI. Think of it as a hidden label. Social platforms read that label to add their own AI badge automatically. Here is what happens to it in our tool.

The hidden label does not survive. Whenever a picture is cropped, given a design, turned or resized to fit a platform, it gets saved again, and saving strips the label out. Video gets rebuilt too whenever captions or cuts are added. That part runs on a separate system we have not checked, so treat video the same way.

There is a second kind of tag woven into the picture itself rather than attached to the file. Google says the one it uses survives being saved again. We do not strip it on purpose, but we have not tested it ourselves, so we are not going to promise it.

So do not wait for Instagram or TikTok to label the picture for you. If it could pass for a real photo, say so yourself. Section 5 covers this.

The same thing in formal legal wording is in our AI transparency notice.

3. What you can switch off

There are four real controls, and one thing you cannot switch off.

Turn off AI images

One setting turns image generation off for your whole account, and you can also skip it for a single batch. With it off, nothing is made up. The tool uses your own product photos first, then pictures you have uploaded, then stock photos. AI still picks which of your photos suits each post, and your text and brand colours may be laid over it, but the picture itself is real. If you do not want made-up pictures on your feed, this is the setting that matters. Ask us and we will switch it on.

Use your own photos and video

Upload from your computer, send from your phone by scanning a code, film straight into our recorder, or build up a clip library we pull from. Posts made from your own footage of your own shop, staff and products are not AI-made at all.

Rewrite or bin anything before it goes out

Every caption can be edited, and every post can be thrown away. Since nothing goes out without your approval, treat the AI text as a first draft. Rewrite it in your own words and it is simply your writing.

Use your own designs

Bring in your own Canva templates and brand designs and have them used exactly as they are, rather than as inspiration for an AI image. Put that together with the setting above and everything people see is your own design work.

What you cannot switch off

There is no setting that stops the AI writing text. Writing captions and ideas is what you are paying for, so a version without it would not be the product. If what you need is no AI at all in your marketing, oriiion is the wrong tool. What we offer is that AI writes the draft and you decide, every single time.

4. What we are responsible for, and what you are

There are two jobs here. We make the tool. You decide what gets posted. That line matters, because it is the part no supplier can take off you.

Ours

Making sure our assistant, chat channels and phone line say they are AI, and giving you a way to reach a person by email.
Naming every company your content goes to, what each does, and keeping that list current.
Never posting to your accounts unless you tell us to, and never treating a draft as approved.
Telling you honestly what our files contain, including when the answer is unhelpful.
Making the switches in section 3 work, and never training AI on your content.

Yours

Reading a draft before it goes out. This is the step that makes everything else work.
Deciding to post, and adding a line when a picture is realistic enough to be taken for a real photo. You published it, so that one is yours.
Choosing your settings, including whether AI images are on or off.
If you set up auto-replies to comments, remember the person getting one is talking to a machine, even though you wrote the words.
Anything you upload yourself. We cannot tell whether a file you supply was made by AI elsewhere, so we treat it as a real photograph — marking it is your call, in the gallery.

We are responsible for how the content gets made and for being straight with you about it. You are responsible for what you post. Be wary of any supplier who says they can take that second half off your hands.

5. What Instagram, TikTok and the others do

The platforms have their own AI labels, separate from the law. They work in two ways: they try to spot AI automatically by reading the hidden label in the file, and they give you a switch to declare it yourself when you post.

Platform	How it labels AI content
Instagram, Facebook, Threads	Adds an AI label it detects automatically from the file, and also gives you a switch to declare AI content. Meta asks you to declare realistic AI video and audio.
TikTok	Adds an AI label, reads the hidden tag when a file has one, and gives you a switch when you upload.
YouTube	Asks when you upload whether the content is altered or AI-made, and shows a note on the video or in the description.
LinkedIn	Reads the hidden tag when it is there and shows that on the picture.
Pinterest, Snapchat, X	Each has its own rules and switches, and they change more often than this page does. Check when you post.

The one thing worth remembering

Automatic spotting relies on that hidden label, and as section 2 explains, it does not survive our tool. So do not wait for a platform to flag our pictures for you, because usually it will not. If you post an AI picture that could pass for a real photo, either tick the platform's own AI box, or just write a short line in the caption saying the image was made with AI. The caption works everywhere, readers understand it, and it does not depend on any platform feature. It takes seconds and it is the most useful thing you can do.

6. Read the law yourself

The original sources, so you can check for yourself. This page is information, not legal advice.

The EU AI Act, Regulation (EU) 2024/1689 — the full official text. Article 50 is the part about telling people when something is AI.
Article 50 on its own — just the one article, easier to read than the whole law.
The European Commission's guidance — the EU's own explanation of how Article 50 is meant to work in practice.
The European Commission's questions and answers — short official answers, including what counts as a person checking a text.

7. Documents and contact

Questions about how we use AI, which company handled a particular request, or about one specific file, go to data@oriiion.ai.

If you would rather deal with a person than the assistant, write to support@oriiion.ai.