Responsible AI policy

1. Principles

• Safe by default. Safety checks run on every account and every plan. No setting or purchase switches them off.
• Explained, not silent. When we refuse a request, we say so and say why.
• Consent for real people. A real person’s likeness is used only with the permission of the person shown, and never for children or public figures.
• Honest about what is synthetic. Every output is marked as made with AI.
• Your inputs are not kept. Files you upload to make something are deleted automatically after six hours.

2. What we refuse

The full list is in the Acceptable Use Policy. Among other things, PrismPoster does not generate:

• nude, sexual or explicit content of anyone, real or generated;
• any content involving a minor in a sexual context, and realistic video of children;
• the likeness of public figures, politicians or other office holders, and face swaps or content that makes a real person appear to say or do something;
• hate, harassment, extremism, self-harm, graphic violence, and instructions for weapons;
• forged documents, scams and impersonation of an authority, company or person;
• music that imitates a named real artist, and named copyrighted characters or brands.

3. Moderation transparency

Before a generation request is processed, its text is checked by an automated AI safety classifier. If the classifier is unavailable, the request is refused rather than let through. The AI services that render your media apply their own safety rules as well.

When a request is refused, PrismPoster tells you, in the app, at the moment it happens:

• which rule it met: when our own policy refuses, the rule is named, for example a public figure’s likeness, violence, a copyrighted character, or a forged document. When the rendering engine declines, we say so and name what usually triggers it;
• why, in plain language, including whether it was your words or an attached picture;
• what to change to make something you can have.

Policy refusals are logged against the account in a moderation record and a compliance audit trail, and serious categories, such as any sexual request involving a minor, are flagged for human review. Continued attempts end the account. Anyone can report content or conduct to safety@prismposter.com, and you can contest any removal, suspension or termination as set out in section 7 of the Terms of Service.

4. Real people and consent

Before a render that uses a picture of a real person, you confirm that you hold the rights and the permission of everyone shown, and we record that confirmation. Photos of children and of public figures are refused. These rules hold on every plan.

5. Marking AI output

Every output carries embedded metadata saying it was made with AI, and C2PA Content Credentials are added wherever our signing service is available. Free accounts also carry a visible label. The details, and your own disclosure duty when you publish, are in our EU AI Act transparency notice.

6. Data and privacy

Files you upload to make something are stored under a temporary area and deleted, with the records that point at them, six hours after upload. Only media you add to a video-editor project is kept. Biometric material is deleted from storage before our database forgets where it was. See our Privacy Policy and security practices.

7. Updates and contact

We update this framework as our product, the law and good practice change. Questions: support@prismposter.com.