Data & AI transparency

Last updated: June 27, 2026
In one line
To generate a question, the text on a sticky note is sent to an external AI model (Anthropic's Claude). PingBack asks for your explicit consent before the first AI analysis, names the model used, never sells your data, and offers an optional anonymization mode that strips common sensitive details before anything leaves our server.
How your data flows
You start an analysis in your browser
From the Miro panel, or by pasting a Miro/Mural board link into the PingBack web app. Nothing is sent until you give AI consent.
PingBack reads the sticky notes
For board links, our server fetches the sticky text from Miro or Mural using the access granted by your OAuth connection (read-only for Mural). It also reads colors, positions, and connectors to understand how stickies are grouped.
Optional: sensitive content is anonymized
If you enabled anonymization, our server redacts common sensitive patterns (see below) before the text is sent anywhere else.
The text is sent to Anthropic Claude
Each sticky (plus a little context from neighboring stickies) is sent to the Claude API, which returns one sharp question per sticky.
A presentation session is stored
The generated questions and minimal metadata are saved in our PostgreSQL database so you can open a shareable presentation view and a QR code. Sessions expire automatically (2 hours by default).
Participants see only the questions
Anyone who scans the QR code can read the questions, react with emoji, and submit anonymous comments. Only the host can edit the session, moderate comments, translate, or export.
Which AI model processes your data
PingBack uses a single AI provider. We do not silently switch models.
Anthropic — Claude (model claude-sonnet-4)
Receives sticky-note text (and, for translations and rubric scores, the generated question text) to produce questions, follow-ups, translations, evaluations, and the optional facilitator summary. PingBack does not control how Anthropic stores API requests. Anthropic privacy policy
Your consent
Explicit, before the first analysis
The first time you ask PingBack to generate questions, a dialog explains that board text will be sent to Anthropic Claude and asks you to agree. No AI call happens until you do. Your choice (and whether anonymization is on) is remembered on your device only — it is never stored on our servers.
Changing your mind. You can re-open the data dialog any time from the Data & privacy link beside the generate button to toggle anonymization, or clear it entirely by clearing your browser storage for this site. The next analysis will ask for consent again.
Anonymization mode
What it redacts
When anonymization is on, our server scrubs these patterns from each sticky (and its neighbor context) before the text is sent to the AI. It runs locally on our server with no extra network call, so it adds no latency:
emails links / URLs phone numbers payment cards CPF / CNPJ @handles
An honest limitation. Anonymization is pattern-based, so it reliably catches well-formed identifiers but cannot recognize every form of sensitive information (for example, a person's name written in free text). Treat it as a strong safety net, not an absolute guarantee. When anonymization is on, the redacted text is also what we store for the session — the original sensitive values never reach the AI or our database.
What we store, and for how long
  • Generated questions, host focus state, emoji reaction counts, and moderated comments are stored in our PostgreSQL database as a presentation session.
  • Sessions expire automatically — 2 hours by default — and expired rows are deleted on a recurring sweep.
  • Pending and rejected comments are never exposed to participants; only the host sees them, and only the host can approve them.
  • We do not keep your raw board after the questions are generated, we do not log board activity for profiling, and we never sell data.
For the full legal detail, retention rights, and how to request deletion, see the Privacy Policy.
Questions about this