AI permissions and recording boundaries

Humanly Cloud lets the writing setup define what the in-platform AI assistant can access and do. The setup can disable the assistant or allow limited discussion, polishing or fuller editing assistance. These permissions govern the configured workspace; they cannot establish that a writer used no external AI service or other source of help.

What do the four AI access modes mean?

Resource and document access follow the selected mode. Read Only does not give the assistant editor access. For an ordinary document in Polish Only, a valid selected passage limits the editing scope to that passage; without a usable selection, document-level polishing is available. When the assistant proposes an edit, the writer decides whether to apply or reject the suggestion.

The setup also offers a web-search choice when AI is enabled and an optional custom AI rule. A custom rule expresses additional task requirements to the assistant. It should not be described as a mathematical guarantee that every generated answer follows an arbitrary instruction. The task's written expectations and the actual interaction record remain relevant to review.

Choose permissions for the task

1. State the allowed assistance in the task instructions: for example, resource discussion, language polishing or broader drafting help.
2. Select the corresponding AI access mode and review any custom rule and web-search setting.
3. Set copy-and-paste and resource-access controls separately. Allowing an assistant and allowing imported text are different choices.
4. Explain the setup to writers before they begin, including what activity and AI interactions the workspace records.
5. Interpret the resulting certificate against those permissions, rather than assuming every AI interaction is a violation.

For example, a task can permit discussion of supplied readings without enabling AI edits to the response. Another task can permit polishing of a selected paragraph. Those tasks have different evidence expectations even if both involve AI chat. See education for the assigned-task workflow and writers for personal writing.

What does Humanly record about AI use?

Humanly records in-platform AI interactions and applied assistance as part of the writing evidence. A conversation, an offered edit and a change actually incorporated into the document are different events. Source composition is about recorded text attribution, not a probability that an AI wrote the whole document.

Certificate readers see the evidence permitted by the certificate's display settings. The evidence guide explains why a final-text percentage can differ from process input volume and why the integrity seal does not turn a recorded action into proof of independent thought.

What remains outside the boundary?

Turning AI access off in Humanly does not disable other applications or devices. A writer can consult external text and type it into the editor. A visibility event can show that the workspace was hidden, but it cannot identify what the writer did elsewhere. Do not translate “no recorded in-platform AI assistance” into “no AI assistance occurred.”

AI use may involve processing prompts, documents and related content through service providers, as described in the privacy policy. For confidential review work, also consult the peer-review guide and the venue's own requirements. Published detector results and their narrower experimental scope are described in the research guide.

Reviewed: 2026-09-24.

Applies to: Humanly Cloud release baseline 75eba744, checked against source revision d5818c71 on 2026-09-24. Product descriptions cover capabilities shared by that release and source, not a new deployment. Available controls and evidence depend on the writing setup. Research findings refer to arXiv:2607.21758v1.

Canonical: https://cloud.writehumanly.net/reference/ai-policy

Markdown version · All reference topics