AI statement
SocialNow OS creates text, images and film with AI models. This statement says where that happens, which models we use, what does and does not happen to your input, and who is responsible for what finally gets published. The Dutch version is the binding one.
Where AI sits in the OS
Milo is the assistant in the OS. It reads what is in your workspace and answers questions, proposes tasks and prepares actions.
The Studio creates posts: copy, image and film, based on your brand passport and the brief you give.
The brand scan reads a website and derives colours, typefaces and tone from it.
The CRM and the calendar use AI for summaries, draft messages and suggestions for the best moment to post.
Which models, and where they run
We use models from Google (Gemini), OpenAI, Anthropic and Higgsfield. Which model performs a task can differ per feature and per moment; we choose on quality, speed and cost.
The current list of providers, with where they process and on what basis data leaves the EEA, is in Annex C of the data processing agreement.
We do not build models and we do not train models. We use existing models and give them context from your workspace.
What happens to your input
What you enter, or what comes from your connected sources, goes to a model only to carry out your instruction, and only the part needed for it.
With every model provider it is contractually excluded that your input is used to train their models. We do not use your data to train or fine-tune a model either.
We do use a knowledge layer: texts and examples with which we ground a model so the answer fits SocialNow. That layer contains no customer data.
Keys you enter for Odoo, Meta, Google or your social channels never go to a model.
Disclosure of AI content
Article 50 of the AI Act (Regulation (EU) 2024/1689) requires those who create synthetic content to disclose it. We implement that as follows.
In the OS it is always visible that a post was created by a model: the Studio shows it while creating, and a post carries that origin with it in your media.
Images and film created with a model carry the provenance data the provider attaches, including C2PA data where the model supplies it. We do not strip it.
If you publish a post on a channel with its own AI label, such as Meta or TikTok, setting that label is up to you. We cannot enforce it for you, and we do not quietly switch it off.
What AI in the OS does not do
Nothing is published on your behalf without you or a team member with write rights clicking on it. A model proposes; a person decides.
No decision is taken about a person that produces legal effects or similarly significantly affects them. There is no automated decision-making within the meaning of Article 22 GDPR.
There is no scoring, ranking or profiling on characteristics of people beyond what you record in your own CRM.
One client's data is never used to produce an answer for another client. A workspace is a boundary, for the model too.
What a model can get wrong
Models sometimes invent something that looks right. They mix up names and numbers, they fill an empty space, and they are just as confident when they are wrong.
So: check every text and every figure before you publish, certainly for prices, claims, names and commitments. The OS does not put an amount or a promise into a post that it did not take from your source, but a model asked for a text may still add a sentence.
Responsibility for what finally goes out under your name lies with you. That is also stated in the terms of service.
Copyright and image rights
Images created with a model on your brief may be used as agreed in the terms of service. The legal position of AI imagery is in motion; we give no guarantee that copyright subsists in such imagery, and therefore none that you can stop others from making something similar.
If you supply your own images, for example photos of your team or products, you declare that you are allowed to use them and that the people shown have consented. We use your images only for your own posts.
Do not ask a model to reproduce the work or the face of an existing person or brand. That is an infringement, even when a model simply does it.
Questions and objection
If you want to know which model produced a particular post, or you want a certain kind of processing to stop: privacy@socialnow.nl. We can switch features off per workspace.