AI transparency

1. How artificial intelligence is used

The oriiion service uses general purpose artificial intelligence models supplied by third parties. We do not train our own models. The models are used to generate captions, scripts, articles, images, video and speech, to analyse material you provide such as photographs and website content, to answer questions in the assistant, and to produce recommendations about what to publish and when.

Get Orion AI AB is a deployer of these systems, and is a provider of the artificial intelligence system that the service itself constitutes when it is made available to you. We are not the provider of the underlying models.

The suppliers of the underlying models are named, with the purpose of each, on our sub-processor page.

2. You are interacting with an artificial intelligence system

Article 50(1) of the Artificial Intelligence Act requires that a natural person interacting with an artificial intelligence system be informed of that fact, unless it is obvious to a reasonably well-informed and observant person from the circumstances. The following parts of the service are operated by an artificial intelligence system rather than by a person:

• The assistant in the web application and the mobile application.
• The assistant when reached over Telegram, WhatsApp, Messenger or SMS.
• The telephone service, in both inbound and outbound calls. Replies are produced by an artificial intelligence system and spoken in a synthetic voice.

A person can ask at any point to be put in contact with a member of our staff instead, and a request to that effect is honoured. Nothing in the service is intended to give the impression that an artificial intelligence system is a person.

3. Content generated or manipulated by artificial intelligence

Text, images, video and audio produced through the service are artificially generated. Where you use the service to alter existing material, for example by editing a photograph or generating video from a still image, the result is artificially manipulated.

Article 50(2) of the Artificial Intelligence Act places the duty to mark synthetic image, audio and video output in a machine-readable format on the provider of the generative model. We are not that provider, and we do not add marking of our own. Two kinds of marking behave differently once a file has passed through our service. Marking carried in file metadata, such as C2PA content credentials, does not survive our image processing: images are re-encoded whenever they are cropped, composed into a design, rotated or resized for a platform, and that discards the metadata block. Video is re-rendered by our editing worker whenever captions or cuts are applied. That worker is operated separately from the application and we have not verified what it preserves, so the same assumption should be made for video. Marking carried in the image or audio signal itself, such as the SynthID watermark that Google states it applies to output from its Imagen, Gemini and Veo models, is according to Google designed to survive re-encoding and ordinary editing. We do not remove it deliberately, but we have not independently verified that it survives our processing and we do not warrant that it does. You should not rely on any marking being present in a file you download from us, and you should not treat its absence as evidence that a file was not artificially generated.

4. What you remain responsible for

Generated content is a proposal. Nothing is published until you decide to publish it. You are responsible for reviewing content before publication and for the content once published, in particular for the following:

• Accuracy. Artificial intelligence models produce confident text that is sometimes wrong, including about prices, dates, availability, legal and medical questions, and the properties of your own products.
• Statements about people. Generated text can assert things about a named person that are untrue, which may be defamatory and, where the person is identifiable, may also be inaccurate personal data.
• Third party rights. Generated images and text may resemble existing works or contain marks belonging to others.
• Platform and advertising rules. The platforms you publish to, and advertising law in your market, impose requirements that generated content does not automatically meet.

We do not warrant that generated content is accurate, original, or fit for any particular purpose. The limitations and allocations of liability in our terms of service apply.

5. Automated decisions and personal data

The service produces recommendations, including suggested publication times, suggested content and suggested advertising audiences. These are presented to you as proposals. Publishing content and committing advertising spend require a decision by you.

Because a person decides, this processing does not amount to a decision based solely on automated processing that produces legal effects concerning you or similarly significantly affects you within the meaning of Article 22 of the General Data Protection Regulation. If we introduce a feature that acts without a human decision, we will say so before it is enabled and will provide the safeguards Article 22 requires.

Content you submit is transmitted to our model suppliers, and personal data within it is transmitted with it. This is described in section 4 of our privacy policy.

6. Prohibited uses

The Artificial Intelligence Act prohibits certain practices outright. You may not use the service to generate or publish material that manipulates a person's behaviour in a way likely to cause significant harm, that exploits vulnerability arising from age, disability or social or economic situation, that infers emotions in the workplace or in education, or that produces social scoring of individuals.

You may also not use the service to impersonate a real person without that person's authorisation, to create sexual or intimate imagery of an identifiable person, or to produce material that is unlawful in the market in which you publish. These restrictions are repeated in our terms of service, and breach of them entitles us to suspend the account.

7. Questions

For questions about the use of artificial intelligence in the service, about which supplier processed a particular request, or about the marking of a particular generated file, write to data@oriiion.ai.