AI disclosure, Holly and platform rules · 11 Proof and guardrails

Holly is the centre of pitch and demo; the trust ladder (H4, H5) is the headline.
A human decision on every LinkedIn post (H6). If asked why: LinkedIn's terms are reported to restrict automated posting; LinkedIn's "AI slop" button punishes generic posts (S1); and the EU AI Act has two duties that matter here. A deployer's duty to label AI-written text applies only to text published to inform the public on matters of public interest, and it falls away where a person reviewed it and someone holds editorial responsibility (Article 50(4)). A separate provider duty to mark AI output in a machine-readable way (Article 50(2)) is being assessed by our lawyer. Never say human approval makes posts AI Act compliant.
Disclosure: Holly identifies itself as AI (H13). Drafting attribution in product copy reads "Holly, in {name}'s voice". No disclosure feature is promised (H14, D9). EU publishing questions go to John for Karl; terms are UK only.
Say "agent" sparingly; prefer "assistant" or "colleague".
A voice built from AI-written posts reproduces that style. Say so; lean on the interview.
No automated engagement, ever. LinkedIn's User Agreement is reported to prohibit bots and automated methods to "create, comment on, like, share, or re-share posts", and LinkedIn has sued scraping operations (Proxycurl, resolved July 2025; ProAPIs, agreement in principle February 2026). Never recommend engagement pods or hacks.
Engagement data stays inside Maktura. Other platforms: LinkedIn only until D13; their rules are added here before anyone sells them. Each person connects their own account. Customers' own rules apply on top: Maktura helps a firm apply its policy and regulator rules; it does not override them.