Particle.news
Download on the App Store

EU Puts AI Act Into Force, Giving Regulators New Power Over Chatbots and Large Models

The move lets the European AI Office and national regulators compel labels, machine-readable provenance and technical records to curb deception and systemic risks.

Overview

  • The European Commission began enforcing key parts of the AI Act on August 2, 2026, requiring chatbots and interactive systems to disclose they are AI and obliging AI-generated images, audio, video and some text to carry visible labels plus machine-readable marks.
  • Regulators can now inspect general-purpose AI (GPAI) models, demand technical documentation and training-data summaries, question company staff and order corrective actions when systems break the law.
  • Article 50’s transparency rules carry fines up to €15 million or 3% of global turnover and breaches of banned practices can lead to penalties of up to €35 million or 7% of turnover.
  • The Commission set up enforcement support including a 60-member Scientific Panel and a Lead Scientific Adviser, complaint and whistleblower channels, and made clear recent incidents involving Anthropic and OpenAI have sharpened emphasis on cyber and loss-of-control risks.
  • Major parts of the law remain phased in: bans on non-consensual sexually explicit and child sexual abuse AI take effect December 2, 2026, stricter high-risk obligations start December 2, 2027, and rules for AI in regulated products begin August 2, 2028.