Privacy notice

Who is responsible

The controller is Rafal Sliwinski, ul. Modrzewiowa 16/10, 21-040 Swidnik, Poland. Write to this address for any request below.

What we process and why

• Account: your email address and a hashed password, to create and secure your account and send verification and password-reset emails (GDPR Art. 6(1)(b), contract).
• Questions and answers: what you type into the chat, the answers, the reports, and the sources read, stored in your conversation threads so you can return to them (Art. 6(1)(b)).
• Uploaded documents: if you attach a PDF or text file to a conversation, the text extracted from it and a numeric embedding of each part of that text, so the chat can answer your questions from it (Art. 6(1)(b)). The original file is not kept. The excerpts used for an answer and your questions about them also appear in the technical traces below.
• Daily usage count: how many research turns you used today, to enforce the free daily limit (Art. 6(1)(b)).
• Technical traces: a record of each conversation turn (your message, the sources read, the answer) kept on our server to find and fix errors (Art. 6(1)(f), legitimate interest in a working service). Traces carry your user id.
• Video notes: when a forecast reads a public YouTube video, Gemini's notes on it and the short question the forecast asked about it, so the same video is not processed again for the same question (Art. 6(1)(f), legitimate interest in a working service). The question is written by the forecast's research step from your request. Video notes are not linked to your account or your conversation.
• Server logs and backups: IP address and request details in web-server logs, and database backups, for security and to recover from failures (Art. 6(1)(f)).
• Blog subscription: if you write to blog@politicalagent.eu to subscribe, your email address, to send you new posts (Art. 6(1)(a), consent, which you can withdraw by writing to the same address).

We do not use analytics, advertising or tracking cookies, and we load no fonts or scripts from third parties.

Blog posts are AI-generated. When you use the chat, an AI system writes the answers.

Cookies and browser storage

We set only cookies that the service needs: a session cookie that keeps you signed in and a CSRF cookie that protects forms. The chat stores the id of your current conversation in your browser's local storage. None of these is used for tracking.

Who receives data

• OpenAI (language model provider) receives your message, the earlier messages of the same conversation, and the text of the sources read, to write the answer. When you attach a document, OpenAI also receives its whole text, to compute the embeddings, and the excerpts of it that answer each question. Data may be processed in the United States.
• OpenRouter (model routing service, United States) receives, only when you ask for a forecast, the forecast's question — which for a follow-up message may include context rewritten from the earlier messages of the conversation — the claims being voted on, and the passages quoted for them from the sources read, and passes them to the providers of the AI models on the forecasting council: free models, and a paid model in place of a free one that fails to answer. Those providers may be located outside the EEA and the United States and may log these inputs, and the free models' providers may use them to improve their models.
• Google (Gemini API) receives, only when you ask for a forecast, the question above (with the same rewritten context) for one council seat, and the address of any public YouTube video the forecast reads, with a short question about it. Google may use these inputs to improve its products. Data may be processed in the United States.
• Brave Software (search provider) receives the text of your question, shortened if long, together with the list of sites to search. Data may be processed in the United States.
• OVHcloud hosts the server, stores off-site backups in its Paris region, and delivers our email.

Transfers outside the EEA rely on the EU–US Data Privacy Framework where the recipient is certified, and otherwise on the European Commission's Standard Contractual Clauses.

For the providers behind OpenRouter's models we cannot confirm either safeguard. If you do not accept that, do not ask for a forecast; no other question sends them anything.

How long we keep it

• Your account, conversation threads and usage counts: until your account is deleted. To delete your account and every conversation in it, email us.
• Uploaded documents: until you remove the file from the conversation, delete the conversation, or your account is deleted. Copies in database backups remain until they are replaced (see below).
• Database backups: off-site copies are deleted after 14 days. A copy is also taken on the server before each software release, and the copies of the five most recent releases are kept; data you deleted can remain in these copies until they are replaced.
• Technical traces: until they are cleared manually. Tell us the date and wording of a question if you want its trace removed.
• Video notes: 30 days after they are written, and then until the database backups that contain them are replaced (see above). Because they are not linked to accounts, deleting your account does not delete them; tell us the date and wording of a forecast if you want its notes removed sooner.
• Server logs: until the next software release, when the server's containers are replaced.
• Blog subscription: until you unsubscribe.

Your rights

You can ask for access to your data, correction, deletion, restriction of processing, a portable copy, and you can object to processing based on legitimate interest. You can also complain to the Polish supervisory authority, the President of the Personal Data Protection Office (Prezes Urzędu Ochrony Danych Osobowych, ul. Stawki 2, 00-193 Warszawa, uodo.gov.pl).