1. Where AI is used
AI does not turn an unavailable capability into an available one and does not prove that an external action succeeded.
- Understanding the post, comment, message, language, intent, and safety context.
- Drafting public replies and continuing approved Setter conversations for the active Page and programme.
- Turning audience questions into content ideas, scripts, plans, captions, and Calendar drafts.
- Guiding a Studio project, selecting suitable source material, and helping generate, edit, narrate, subtitle, or assemble media.
- Summarising workspace knowledge and showing operational recommendations.
2. When a person talks directly to JoyReply
Where the law requires it, the first automated interaction or the surrounding interface identifies that the person is dealing with an AI assistant acting for the named creator or brand. The notice does not need to be repeated in every message when the identity remains clear.
Before 2 August 2026, a controlled private workspace may use the documented pre-deadline setting. From 2 August 2026, JoyReply's current disclosure rule requires the notice regardless of that setting. The workspace may customise the brand wording, but it must not pretend the assistant is the author personally.
3. Setter boundaries
Setter can explain approved programme information, ask relevant questions, and provide the saved booking link. It cannot confirm a meeting, time, payment, discount, accreditation, or outcome without a reliable external source.
A clear stop or no-contact request is terminal. If the conversation becomes unsafe, unsupported, or impossible to answer reliably, automation stops instead of continuing indefinitely.
4. Generated and edited media
JoyReply records available provenance about source files, generated or edited segments, voice origin, model or workflow revision, review state, destination, and the disclosure decision. Machine-readable provenance is preserved or added where the workflow supports it.
Not every AI-assisted edit needs a visible watermark. Realistic synthetic scenes, cloned real-person faces or voices, fabricated events, deepfakes, and other material that could be mistaken for authentic content may require a clear visible label and the destination platform's AI information control. The owner must review that decision before publication.
5. What remains under user control
JoyReply shows the active workspace mode, capability gates, stop state, review state, Page selection, and external confirmation boundary. The owner remains responsible for choosing approved sources, checking sensitive claims, approving content, selecting the destination, and complying with the platform's disclosure tools.
6. Limitations and questions
AI can be confidently wrong, misunderstand Romanian phrasing, miss context, or produce unsuitable media. Report an incorrect or misleading result at contact@joyreply.com. For questions about conversation data or provenance records, use privacy@joyreply.com.
The practical rule
Disclose the AI interaction when required, never pose as the author, stop on request, label realistic synthetic media when needed, and keep provenance even when a visible label is unnecessary.