AI processing

What AI does here

On this platform, AI only produces suggestions and translations. No decision about your admission, membership tier, price or access is made automatically; a person makes every decision.

The processor and where it sits

Our processor is Google (Gemini API) or Anthropic (Claude API) — whichever is active in the system settings — and both process outside Türkiye. That is a cross-border transfer, and the cross-border transfer page sets out the basis and the recipients. The data is not used to train models.

What data is sent

• The CV and LinkedIn profile PDF you upload yourself — only with a separate consent.
• Your intake voice message — only until it is transcribed, and only under that same consent.
• The answers you write to the intake questions, to propose tags and markets.
• The photo of a business card scanned at an exhibition, to read the name and company.
• Public company information, to build a company dossier and research exhibitors.
• Chat message text and form text, for live translation across the five languages.
• The need description and professional details, to write the match rationale in matchmaking.

Voice: what happens and what does not

Audio transcription goes through Gemini only, and audio is sent for transcription alone; an explicit instruction forbids identifying or describing the speaker. No voice print, speaker trait or audio vector is stored — we collect no biometric data. The audio file is deleted immediately after a successful transcription.

What we deliberately do not do

• No LinkedIn page is read or scraped; the supported route is your own exported PDF.
• Any suggestion touching health, religion, political opinion, ethnicity, sexual orientation, union membership or criminal record is discarded before it is stored.
• No model output lands on a profile automatically, and no behavioural flag affects an invitation until a person has reviewed it.
• No behavioural score is produced by AI; the membership score comes from transparent, numeric rules.

AI output is shown separately

Anything the model produced is labelled as a public impression and sits in its own frame, never beside confirmed data. A database constraint holds that boundary, not code discipline.

How long it stays

• Audio file: deleted immediately after a successful transcription, and in any case within seven days.
• CV and PDF: until the membership is deleted, at most twelve months.
• Suggestions you never accepted: until consent is withdrawn or they are deleted by hand.
• The request and response text at the processor: it is not used to train models, and that service's own data-processing terms set how long it is kept, not us. What is in our hands is that the file and the text leave our own system within the periods above.

Your right

You can withhold this consent and still be a full member: you then fill in your profile by hand. Whenever you withdraw it, open work is cancelled and the files and pending suggestions are deleted. Write to ⁨privacy@meetupbusinessclub.com⁩ with any question.