Guides · Article 50 · EU AI Act
AI meeting transcription and the EU AI Act
AI notetakers and meeting-transcription assistants record calls, produce transcripts, and summarise action items — and some infer sentiment or engagement. More than one Article 50 paragraph can come into view depending on what the tool actually does. This guide maps the possibilities from the regulation text. It is informational only, not legal advice, and does not name any tool category as in or out of scope.
Transcription, summaries and the Article 50(2) carve-outs
Article 50(2) asks providers of AI systems generating synthetic text to mark outputs in a machine-readable format and detectable as artificially generated or manipulated. A verbatim transcript raises a threshold question: the paragraph does not apply to the extent the AI systems perform an assistive function for standard editing or do not substantially alter the input data provided by the deployer or the semantics thereof. Straight speech-to-text arguably sits close to that language, while an AI-written summary that rephrases and condenses may look more like generated text. Whether a transcript or a summary falls inside 50(2), and whether a carve-out applies, are fact-specific questions for the provider of the system to work through with counsel.
Sentiment and engagement scoring — Article 50(3)
If a transcription tool goes further and infers participants' emotions from voice or text, Article 50(3) can come into view. That paragraph is written on the deployer of an emotion recognition or biometric categorisation system: they must inform the natural persons exposed of the operation of the system, and process the personal data in accordance with EU data-protection law. Whether a given sentiment feature is an 'emotion recognition system' as defined in the Regulation is fact-specific, and this duty should be coordinated with your data-protection documentation rather than handled alone.
A note on the meeting bot itself
Where the tool joins a call as a participant that people interact with, a separate Article 50(1) question can arise for the provider of that system about informing people they are interacting with an AI system, subject to the 'obvious' exception. Mapping which paragraphs are in play — and whether you are provider-side or deployer-side for each — is the useful first step. Our free scope check gives an informational read; confirm with qualified counsel.
Common questions
Does an AI notetaker need to disclose anything under Article 50?
It depends on what it does. Transcript and summary output raises provider-side Article 50(2) questions with editing carve-outs; sentiment scoring can raise deployer-side Article 50(3) notice under emotion recognition; a bot people interact with can raise Article 50(1). Each is fact-specific — verify with qualified counsel.
See what may apply to your business
Answer seven quick questions for an automated, informational indication of which Article 50 obligations appear likely to apply — free, and not legal advice.