Privacy Policy
AI Tumor Board is an open-source research demo. Privacy is built into the architecture, not bolted on. Here's exactly what happens to the case text you paste.
What we do NOT do
- No user accounts. There is no login, no email signup, no profile.
- No persistent storage. We don't keep a database of your cases. The text you paste is held in server memory only for the duration of the board run (typically 2–5 minutes), then dropped when the session expires (30-minute TTL).
- No analytics, no tracking, no cookies. We do not load Google Analytics, Mixpanel, Segment, Facebook Pixel, or any other third-party tracker.
- No advertising, no data sales. Your data is never sold, rented, licensed, or shared with third-party marketers, data brokers, or advertisers.
- No training on your data by us. We do not collect cases to fine-tune models or build datasets.
What we DO send to third parties (for the app to work)
To produce a board recommendation, the case text and the agents' tool queries are sent to the following services in real time. We do not control these providers' retention policies — each one's terms apply.
- Google (Gemini) — model inference: the case text is included in the prompts sent to Google's Gemini API so the specialist agents can reason about it. Google's API data-usage policy applies — see ai.google.dev/gemini-api/terms. As of writing, Google states that paid-tier Gemini API inputs are not used to train models by default.
- NCBI / PubMed (Entrez API): receives the search queries the agents construct (e.g., "esophageal adenocarcinoma neoadjuvant chemoradiation"), not the full case. NCBI logs queries per their NLM policy.
- Europe PMC, Semantic Scholar, ClinicalTrials.gov, openFDA, DailyMed, RxNorm, CIViC: each receives the specific queries the agents make. Each has its own data policy.
- Brave Search: receives web-search queries if an agent invokes the fallback search tool.
None of these services receive your name, email, IP-mapped identity, or any other personal identifier from us — only the query text and the case text (the latter only goes to the LLM provider).
⚠️ Please remove PHI before pasting cases
Even though we don't store your input, third-party services (especially the LLM provider) will receive whatever you paste. Please de-identify cases before submission, removing the 18 identifiers listed by HHS at hhs.gov — names, dates of birth, MRNs, addresses, photos, etc.
Open source
The full source code is at github.com/Roupen92/AI-tumor-board. You can audit exactly what data is sent where, run the app yourself, or fork it. There is no hidden server-side tracking layer — what you see in the repo is what runs.
Questions or concerns
Open an issue at the GitHub repo or email the author: roupen.odabashian@gmail.com.