State of M&A Data Rooms — Q1 2026 Read the report →
Peony LogoPeony

Sharing AI Artifacts (Live HTML) Securely

Upload an .html/.htm AI artifact from Claude or GPT to a Peony data room and share it live — JavaScript runs for the viewer, under watermark, screenshot protection, NDA, permissions, and analytics.

Last updated June 18, 2026

Sharing AI Artifacts (Live HTML) Securely

This page covers how to share an AI-generated HTML artifact — a dashboard, calculator, financial model, or mini-app built in Claude, ChatGPT, or any tool — through a Peony data room so it runs live for the recipient while staying under a full security layer. Peony renders .html and .htm natively in the in-browser viewer with JavaScript executing, so the artifact works as intended instead of being flattened into a dead PDF or a screenshot.

If you have an interactive work product you built with an LLM and you need to send it to investors, buyers, lenders, or a board without losing control of who opens it, this is the workflow. For the commercial overview and use cases, see Share AI-generated documents securely.

What is an AI artifact, and does Peony render it live?

An AI artifact is a self-contained interactive file — usually a single .html file — that an LLM generates for you: a valuation dashboard, a pricing calculator, a scenario model, a one-page mini-app. It has its own HTML, CSS, and JavaScript, so it does real work in the browser rather than just displaying text.

Yes, Peony renders it live. When you upload an .html or .htm file, Peony displays it in its secure in-browser viewer with JavaScript executing, so the artifact runs exactly as it would in a normal browser — sliders move, inputs recalculate, charts redraw. It is not converted to a static image or a PDF first. This is the difference that matters: most data rooms convert HTML to a watermarked PDF before showing it, which kills interactivity and turns a working app into a flat page. Peony keeps the artifact live and wraps it in the same control layer as every other file — per-viewer watermark, screenshot protection, NDA gate, granular permissions, page-level analytics, and instant revoke. More than 5,900 customers share documents on Peony this way.

For the format reference, see Supported File Formats. For the deal-side context, see how to securely share a Claude artifact.

How do I upload and share a live HTML artifact?

Export the artifact as a single .html file, upload it to a data room, layer your security on the link, and send a per-viewer link. The artifact previews live in the browser; the viewer sees the running app inside Peony's viewer rather than a screenshot.

The key is to keep the artifact self-contained. If your LLM produced one HTML file with the CSS and JavaScript inline, it will render cleanly. Save Claude or GPT output as a .html file (most tools have a download or export option, or copy the code into a file ending in .html), then follow the numbered setup below.

Numbered setup: share an AI artifact in 6 steps

  1. Export the artifact as one .html file. From Claude, ChatGPT, or your editor, save the interactive output as a single file ending in .html (or .htm). Inline CSS and JavaScript render most reliably. Keep it under 25 MB.

  2. Create or open a data room. Dashboard → New Data Room (or open an existing one). Name it by the deliverable and audience, for example "Project Atlas — Lender Model".

  3. Upload the .html file. Drag it into the room or use Upload. Peony detects it as HTML and previews it live in the viewer — open it once yourself to confirm JavaScript runs and the artifact behaves as expected.

  4. Set granular permissions. From Permissions, control exactly who can open the artifact and which other files in the room they can see. Create a separate gated link per viewer or per group so each one carries its own trail.

  5. Layer security on the link. In Link Settings, turn on the protections your audience requires — dynamic watermark, screenshot protection, NDA gate, email verification, download prevention, and link expiry. See the next section for which tier unlocks each.

  6. Send the per-viewer link and watch the analytics. Share the gated link. From the Analytics tab you see who opened the artifact, for how long, and how often. Revoke any link the moment access should end.

How do I secure a shared AI artifact?

Layer Peony's controls on the link so the live artifact carries the same protection as a sensitive PDF: a per-viewer dynamic watermark burned into the running page, screenshot protection, an NDA gate, granular permissions, page-level analytics, and instant revoke. Each control sits at a specific plan tier — here is exactly where.

  • Dynamic watermark (Data Room $52/admin/month and up). Each viewer's name, email, and a timestamp are burned directly into the live render, so any screen photo or recording carries the leaker's identity. Watermarks are not available on Business or Free. See Dynamic Watermarks or the watermarks feature page.

  • Screenshot protection (Business $30/admin/month and up; Screenshield on Data Room and up). Desktop screen-capture blocking is included from Business up across major browsers on macOS and Windows. Screenshield, which blocks mobile screen capture on iOS, iPadOS, and Android, is on Data Room and up. See Screenshot Protection.

  • NDA gate (Simple NDA on Business; Advanced NDA on Data Room). Require a viewer to accept an NDA before the artifact opens. Business includes a Simple NDA (view and acknowledge). Data Room adds the Advanced NDA — a digitally signed PDF with countersigning, stored for counsel. See NDA Gates or the NDA feature page.

  • Granular per-file permissions (Data Room and up). Control who sees the artifact versus the rest of the room, and create one gated link per viewer so each has isolated access. Available from the Data Room tier.

  • Page-level analytics (included). The Analytics tab shows who opened the artifact, total time on it, and return visits — useful intelligence about which viewers are seriously engaged. See Page Analytics.

  • Instant revoke. Kill any shared link immediately and the holder loses access the next time they try to open the artifact. To revoke one individual viewer while everyone else keeps theirs — without disturbing anyone else — use granular per-viewer permissions on the Data Room plan ($52/admin/month) and up. See Revoke Access.

Because watermark and screenshot protection are attribution and deterrence layers, treat them as identity-binding on the live render, not as a physical barrier against a determined leaker. The combination — signed NDA, per-viewer watermark, blocked capture, and a complete access log — is what makes a leak traceable and a deal-grade artifact defensible.

Can I connect my own LLM and audit it?

Yes — on the Enterprise plan you can connect your own LLM (GPT, Claude, or Gemini) to the room, and every query the AI runs and every document it accesses is logged in the same access record as a human viewer. Connecting an external LLM is an Enterprise-only capability; lower tiers use Peony's self-contained native AI for document Q&A and extraction.

What makes this useful is the audit, not the connection. The moment an AI assistant reaches confidential documents, your access log normally goes silent on what the AI actually read — leaving an unattributed actor in the room — which is exactly the record counsel or a regulator will ask for. With Peony's Enterprise audit, the connected AI gets a row like every other viewer, so you can produce the complete list of documents it touched and the queries it ran during diligence. For day-to-day document questions on lower tiers, Peony's native AI Q&A answers from the room's own documents without any external connection.

What if my artifact won't render?

If a .html file doesn't render as expected, it is almost always because the artifact pulls assets from external sources or exceeds the preview size limit. Self-contained files with inline CSS and JavaScript render most reliably; keep the file under 25 MB.

Re-export the artifact as a single file with styles and scripts inline rather than linked, confirm the extension is .html or .htm, and open it once yourself in the viewer before sending. If it still won't preview, contact support — and as a fallback, any artifact can be shared as a PDF, though that flattens the interactivity.