Standard
EU package amending digital rules, including the AI Act, to simplify compliance; it delays high-risk AI obligations to December 2027 and August 2028.
EU AI Act high-risk duties, Article 14 human oversight included, now start on 2 December 2027 for Annex III systems, a dev.to analysis says. The duty itself is unchanged, so teams whose agents fall under Annex III still need a person who can override or stop them.
Reality
EU tech chief Henna Virkkunen said on October 9 that the AI Act, as it stands, equips Brussels to contain rogue AI agents. Firms deploying agents in Europe should plan for stricter use of the existing rulebook while its heaviest obligations stay deferred.
Perspective Coverage
4 publishersReality
Article 10's dataset traceability duties stack on top of the GDPR. Location, operator and applicable law each become documentation, and a sovereign region only settles the first of those three.
Reality
MIT's committee says models can credibly do essentially every written assignment it sets, which turns the problem from dishonesty into the missing human interactions that education and training products were built to sell.
Reality
Article 50 and California's transparency act both switched on 2 August 2026. The watermarking work belongs to model vendors; a shipping team owes chatbot, deepfake and public-interest text notices.
Reality
Regulation E asks whether a transfer was authorized. That question has no clean answer when software presses the button, and the consumer is the one who finds out.
Reality