How our AI features work, and their limits.
DRAFT — pending review by qualified legal counsel. The technical description below matches how Raigozen's AI features actually work today; bracketed items need a real decision from the business.
All of these send your input (and, for file features, file contents) to a third-party AI provider (currently Anthropic and/or OpenAI, depending on the feature) to generate a response.
AI-generated text, code, or images can be factually wrong, biased, or unsuitable for your purpose. You're responsible for reviewing AI output before publishing, relying on, or acting on it — including for anything AI-generated on a Website Builder site you publish to real visitors.
If a required provider API key isn't configured, or a provider is down, the affected feature returns a clear error rather than a silently fake response — we never fabricate an AI reply. In a genuine cost or abuse emergency, an administrator can disable AI features platform-wide; if this happens, in-progress work (drafts, chat history) is preserved and unaffected.
In addition to the general Acceptable Use Policy, you may not use Raigozen's AI features to generate: content that impersonates a real person without consent, malware or exploit code, child sexual abuse material, or content designed to harass or defame a specific real individual. We rely on our AI providers' own safety systems as a first line of defense and may additionally act on reports of misuse.
See our Privacy Policy's "AI features" section for what happens to data sent to Anthropic/OpenAI, including links to their own privacy policies.
Questions: [AI/Product Contact Email].
