The EU AI Act (Regulation (EU) 2024/1689) is the first horizontal legal framework for AI in the EU. It classifies AI systems into four risk classes (unacceptable, high, limited, minimal) and imposes graduated obligations per class — from outright bans, through conformity assessments, to pure transparency duties.
For an AI phone assistant the "limited risk" tier typically applies: callers must be able to recognise that they are talking to an AI (Art. 50). That is technically trivial via a short opening announcement, but it must be anchored in the default configuration and documented in the privacy notice.
If the assistant is used for more sensitive purposes (credit scoring, HR decisions, law enforcement) the classification quickly flips to "high risk" with substantially stricter duties: risk management, data governance, technical documentation, human oversight, CE conformity. Violations can be fined up to 7 % of global annual revenue.