Last updated 5 July 2026
Privacy Notice
THEMA Beacon is a private photonics-intelligence hub for THEMA Foundries. This notice explains what personal data it processes and why. It is written to be plain and honest — if the system starts processing anything beyond what is described here, this page is updated first.
Who is responsible
During the current trial, Pierre-Louis Berlemont operates Beacon and is the data controller. If THEMA adopts Beacon, THEMA becomes the controller and PL acts as a processor under a data-processing agreement.
What we process
- Account: your name and email address, and whether you are an admin or member.
- Sign-in sessions: a session token plus the IP address and browser user-agent of your logins (needed to keep you signed in and to secure the account).
- Your content: the chats you have with the research agent and the research documents you generate.
- Tracked entities: which entities you follow and how they were added (followed directly, or connected via a research run), stored with your user id for attribution.
- Usage events: first-party product-analytics events — sign-ins, pages visited, briefs opened, chat turns, research runs, and entities followed — recorded with your user id so we can tell whether Beacon is being used (for example, to notice when someone has gone quiet) and improve it. These stay inside Beacon; they are never sent to a third-party analytics service.
No special-category data is collected, and there is no advertising or third-party analytics. Our emails contain no tracking pixels, and we do not measure whether you open an email. We do record the first-party usage events described above: in particular, when you open a daily brief from the link in its email, the hub logs that visit as a usage event so we can measure how many briefs are actually read.
Where your data lives (EU by default)
Your account, sessions, and content are stored in the EU (Neon in Frankfurt; agent compute on Fly.io in Amsterdam; the hub on Cloudflare's EU edge). Two flows leave the EU by necessity:
- The language model (Anthropic, US): your chat and research prompts are sent to Anthropic to generate answers, under Standard Contractual Clauses and Anthropic's DPA. API data is not used to train models.
- Web search (Tavily/Brave, US): your research queries (not your account data) are sent to fetch public results.
Sub-processors
| Provider | Purpose | Location |
|---|---|---|
| Neon | Primary database | EU — Frankfurt |
| Cloudflare | Hub hosting + embeddings | EU edge |
| Fly.io | Agent compute | EU — Amsterdam |
| Resend | Transactional email (sign-in links, digests) | EU sending |
| Anthropic | LLM processing of chat & research | US — SCCs + DPA |
| Tavily / Brave | Web search for research queries | US |
How long we keep it
- Raw fetched article content: 30 days.
- News items and usage events: 12 months.
- Chat sessions: kept until you delete them, then purged within 30 days.
- Tracked entities: until unfollowed; the attribution to you is unlinked when your account is deleted.
- Encrypted backups: 35 days.
Your rights
You can request a copy of your data (access) or its deletion (erasure) at any time. We respond within one month. Deleting your account removes your profile, sessions, chats, and messages; shared intelligence documents are retained but unlinked from you.
To exercise a right, or for any question, contact pierrelouisberlemont@gmail.com. You also have the right to lodge a complaint with your data-protection authority.
Cookies
Beacon uses one essential cookie to keep you signed in. There are no tracking or advertising cookies, so there is no consent banner.