Artificial Intelligence Policy

Causfy uses artificial intelligence for four things: an assistant that helps you use the app, features that help you draft your cause and generate images for it, automatic moderation of content, which a person can always review, and the identity verification of those who sign with a document and a selfie.

The assistant never sends or publishes anything on its own, and never accesses the data of those who sign.

What artificial intelligence generates may contain errors. You decide whether to use it, and you are answerable for what you publish.

We do not train models with your content.

Words with an initial capital have the meaning given to them by the Terms of Use and the Terms for Creators.

1.1. The assistant. On causfy.app there is a conversational assistant that guides you on how to use the Platform, helps you fill in forms and answers questions about your own causes: how many signatures they have, how your tickets are going. You can talk to it by text or by voice.

1.2. Drafting and image generation. When creating or editing a cause you can ask for help drafting the title, the story and the petition, and generating an image from your instructions.

1.3. Moderation. Every cause is reviewed before publication, and the review combines automated means based on artificial intelligence with human review.

1.4. Using them is voluntary. None of the help features is necessary to create or manage a cause.

1.5. Identity verification of those who sign. At the “verified identity” level, an artificial intelligence system from a specialised provider, which Causfy runs on its own servers, reads the document, compares the face with its photograph and checks that it is a real person. Its sole purpose is to confirm that the person is who they say they are. It operates in the European Union, in an isolated environment with no internet connection, so that neither the provider nor any third party receives your images. It is used only with your explicit consent and does not keep images or templates. The model is not trained with the data of those who sign. If the result is negative, you can try again or sign through another channel, and ask for a person to review the decision.

2.1. What it does. It guides you, prepares forms and reads your own data: your causes, the number of signatures per day and your support tickets. It never sends or publishes anything on its own: everything it prepares is confirmed by you.

2.2. What it does not do. It never accesses the data of those who sign your causes: only the counts.

2.3. Where it runs. The assistant’s model runs on Causfy servers at OVH (Germany). If it moves to another provider, this Policy will be updated beforehand.

2.4. Your conversations are not stored. They live in memory for as long as the session lasts and disappear when it is closed.

2.5. Voice. If you speak to it, the microphone audio is transcribed on those same servers and is not stored. For the spoken reply, the text of the reply is sent to a speech synthesis provider in the United States; your voice is never sent to it.

2.6. Your own assistant, connected. In addition to Causfy’s assistant, you can connect to your account the one you already use via the MCP protocol, with the permissions you grant it; today, Claude, by Anthropic. That assistant is not Causfy’s: what it reads from your account is sent to the company that provides it, under its own terms, and what it does from your account is done by you. It never accesses the personal data of those who sign: only counts and what is public. You can revoke the connection whenever you want from the app.

3.1. The output of these features is offered as help, as is, and without any warranty of any kind.

3.2. It may be inaccurate, incomplete or inappropriate. It is not legal, tax, medical, financial or any other professional advice.

3.3. The Platform notifies you when content has been generated with these features.

3.4. The output does not express Causfy’s position and does not bind it to anything.

4.1. You are answerable for what you enter into these features and for what you decide to publish with them.

4.2. Review the output before publishing it: that it is accurate, that it is lawful and that it says what you mean to say.

4.3. Content generated with these features is subject to the Terms and to the Platform Rules like any other. That a machine generated it does not excuse any breach.

4.4. You may not use these features to impersonate a person or entity, to mislead about facts, people or results, or to circumvent the prohibitions of the Terms.

5.1. What it reviews. The text of the cause, including the name and position of its Addressees, and its images, to check that they comply with the Platform Rules.

5.2. Always with a person behind it. When the automated result is not clear, a person decides. And if your cause is rejected, suspended or removed, you have the right to ask for a person to review the decision, in accordance with the moderation section of the Terms.

5.3. No measure concerning a cause or an account is adopted in an exclusively automated manner without that possibility of human review.

6.1. The drafting, image generation and moderation features process the content of your cause: the title, the story, the petition, your instructions for the image, and the name and position of the Addressees, who are third parties. That content is sent to Microsoft Azure, which processes it on behalf of Causfy in the European Union.

6.2. Do not enter into these features other personal data of third parties, or access credentials, payment details or identity documents.

6.3. The assistant processes your messages and what its tools read (your causes, counts and tickets), on Causfy servers. Never the data of those who sign.

6.4. Causfy does not train its own models with Users’ content, and its providers may not use it to train theirs.

6.5. The details of providers, locations and retention periods are in the Privacy Policy.

7.1. You may not use the Platform, or what you generate with these features, to develop, train or improve artificial intelligence models or systems.

7.2. Automated access to these features is not permitted, nor is using them outside the Platform’s interface.

7.3. Their use may be subject to limits per User and per period.

8.1. We do not guarantee that these features will always be available, and we may modify, limit or withdraw them.

8.2. Providers and models may change. If the change affects where your data is processed, it will be reflected in the Privacy Policy.

8.3. This Policy may change. The date of the last review is in the header.

Questions about this Policy, and those relating to the processing of data, are sent to:

• Questions about this Policy:
• Data protection:

The details of the party responsible are set out in the Legal Notice.