AI and narration disclosure
AI and text-to-speech are optional tools with deliberately bounded context. They can help, but their output is not guaranteed to be correct.
AI starts only when you ask
FrenSpace does not call AI on page load or continuously scan your work. A provider request starts only after an explicit action such as sending an Assistant message, asking a shared document question, asking Group AI, or generating narration.
Context depends on the feature
- Personal Assistant: uses bounded personal tasks, projects, calendar events, and document metadata. It does not currently read extracted personal document text or narration audio.
- Shared-document Q&A: is grounded only to bounded extracted text from the selected shared document and requires active membership.
- Group AI: uses bounded, authorized shared projects, tasks, chat messages, confirmed decisions, and document metadata from one space. It does not automatically receive private personal context or document page text.
Personal context must not automatically enter a shared AI request. Context from another shared space must not enter the current space's request.
Human control over decisions
AI takes no part in your shared decision log. It does not create, suggest, promote, confirm, reject, edit, or delete a decision. Every decision is written by a member, and an owner or admin must separately confirm a proposal before it becomes official.
Accuracy, citations, and availability
- AI answers can be incomplete, inaccurate, or unavailable.
- Citations show application sources that were available to that request. A citation is not proof that the source or answer is correct and never grants additional access.
- Verify important outputs against the original material and other reliable sources.
- Per-user quotas, global budgets, cooldowns, and provider availability may limit access.
Providers and test environments
Provider use is controlled by server configuration and server-managed infrastructure credentials, not by a user's personal ChatGPT subscription. Test environments may use deterministic test providers so automated tests do not contact real AI or speech services.
Narration
- Narration is generated only after an explicit request from selected, already-extracted page text. It is not a human recording.
- Current single-page generation is bounded. Multi-page collections require an explicit bounded page selection and browser-driven processing; there is no automatic whole-document narration.
- Generated speech can contain pronunciation, pacing, or segmentation errors.
- Narration may repeat errors from PDF text extraction and is not a replacement for reviewing the source document.
- FrenSpace does not claim that narration is accessibility-certified.