AI transparency and compliance (EU AI Act)
AI labels on outgoing posts, mandatory human approval, a settings lock, data retention and a downloadable compliance report.
Updated:
On Compliance & AI transparency you decide how Sayweek labels posts that came from an AI draft, guarantee human approval and download the compliance report. Disclosure is off by default. The page is information, not legal advice – please ask a lawyer what applies to you.
Set up AI disclosure
- Open Settings → Compliance & AI transparency.
- Under Default AI disclosure, choose the Disclosure: None, Subtle label or Hashtag (#AIassisted).
- Choose the Label position: End of the text (before the hashtags) or First comment (where available).
- Optionally enter a Custom label text; the Preview shows how it will look.
- Under Per brand and channel, override the default for a brand or a channel.
The disclosure is added to the outgoing copy only (never to the saved post), and only when the post came from an AI draft. Platform-native AI label sets TikTok's and YouTube's own label on videos – switch it on only when the video itself was made or altered with AI.

Human approval and lock
With Every post is approved by a person on, any automated approval (API, rules) is refused. After Lock these settings, only owners can change the compliance settings and human approval can't be switched off. Both are owner-only; while unlocked, editors can change the disclosure too.
Data retention
In Auto-delete ideas you can choose 30, 90, 180 or 365 days (owners only). After that, idea box items, mail, calendar and CRM ideas and AI topic suggestions are deleted; posts are not affected.
Compliance report
- Under Data export and report, set the period in From and To.
- Pick a brand, or keep All brands.
- Click Download CSV or Download PDF.
The report lists every post sent in the period: AI involvement, human change, approver and time, AI disclosure.
Was this article helpful?
Related articles
- How does the AI use your data?What Claude receives to write a post, what stays with us, and why the models don't train on your data.
- Approving postsNothing goes out without approval. How to approve a post, how two-step approval works, and how to answer a client's comment.
- Roles and permissionsOwner, editor, contributor and viewer: who can see and change what in the workspace.
- Activity log: who did whatOwners see approvals, deletions, settings changes, sign-ins and other security events in the activity log.