AI & transcription: consent and data protection

From what is said in a programme item, iveo can use AI to produce a summary, a cleaned-up transcript and the key quotes. This page explains why every speaker has to consent to that, how iveo collects the consent, and where the AI processes your data.

Why consent matters

An AI analysis works with what people say on stage. Its results (summary, cleaned-up transcript, quotes) can be attributed to individual people, which makes them personal data. That is why iveo asks every single person for consent before their contributions go into an AI analysis with their name. Without consent, a contribution is only included without a name, as are remarks from the audience.

Every decision is stored together with its timestamp and the version of the consent text the person agreed to. As organisers, you can therefore show at any time who said yes or no to what, and when.

Where the AI processes your data

The AI analysis uses the Claude model by Anthropic, accessed via Amazon Web Services (AWS Bedrock):

• Frankfurt data center (origin region)
• EU only
• No transfer to the US
• No training on your data

Requests start in the Frankfurt data center and may be distributed to other AWS data centers within the EU. They never leave EU territory. Your data is not used to train AI models. iveo uses the results exclusively for the event in question.

How consent is collected

Asking: Organisers and producers collect consent on the programme item, in the “Media assets” card under the “Transcript” tab. It shows how many speakers have agreed, and “Collect consent” opens a pop-up with every confirmed speaker and their status. From there you send the request, remind individuals with “Remind” or write to everyone still open at once. The same person can be asked again after 24 hours at the earliest.

Consent on someone’s behalf: If a person gave you their consent by email or in conversation, you record it for them in the same pop-up and confirm that with a checkbox. The person receives a short email with a link to change their decision and sees under “My settings › Privacy” who recorded the consent and when. You can withdraw a consent recorded this way; a person’s own decision can only be changed by them.

Answering: Speakers receive an email from iveo; replies to that email go straight to the event team. Using the link in the email, they agree or decline, no login needed. Placeholder addresses are never contacted.

Other routes: The invitation to iveo also contains a section on the AI analysis, and a pop-up asks the same question on the first visit to iveo.

Status and changes: You can see where things stand in the programme item’s “Transcript” tab. Speakers also answer the question there directly with yes or no and can change their decision at any time under “My settings › Privacy”.

The decision applies per event, to all of the person’s programmes in that event. You activate the AI analysis of a programme item with “Activate AI use” once at least one person has agreed; once every confirmed speaker has agreed, it happens automatically. “Don’t use AI for this programme” keeps a programme item free of AI analysis altogether.

Who sees what

• Organisers and producers of a programme see the consent status of that programme’s speakers.
• Directors and moderators see the status read-only.
• Speakers only see their own status.

Declining and withdrawing

From the original transcript iveo creates three AI versions: “Summary”, “Cleaned up” and “Best quotes”, each marked as AI text. Whoever has consented is named in them; whoever declines or later withdraws consent stays included without a name. After a withdrawal iveo recreates the version. The original transcript is not affected.

GDPR: access and email links

• Access: your own consent decisions are part of the self-service export in iveo, described on the Data protection and hosting page.
• Links in emails are valid for 90 days. iveo stores them only in encrypted form, as a checksum, and deletes them afterwards.

Recording & transcription (coming soon)

Recording and transcribing programme items is in preparation and not yet available. The plan is for transcription to run locally on the control-room Mac using the Whisper speech-recognition model, with the transcript and recordings uploaded to iveo in Frankfurt only afterwards. We will update this page once it is ready.

Questions about the AI analysis or consent? Write to us at hello@my-iveo.de.