Transparency obligations under Article 50 of the European Union’s Artificial Intelligence Act began to apply on 2 August, requiring providers to disclose when a person is interacting with an AI system and to label AI-generated audio, images, video and text, including deepfakes. The obligations took effect on schedule even though the EU’s wider AI Act timetable was reshaped in July by the Digital Omnibus on AI, which pushed back compliance deadlines for higher-risk AI systems by 16 months to December 2027, and for AI embedded in regulated products such as medical devices to August 2028.
What changed, and what did not
National authorities and the EU’s AI Office gained the power on 2 August to fine companies up to €15 million or 3% of global annual turnover for breaches of the transparency rules, a meaningful enforcement threat for any company operating in the EU market. A narrower carve-out applies to watermarking requirements for AI content already on the market before 2 August, which have a grace period until December. The high-risk system rules, covering uses such as biometric identification, employment screening and critical infrastructure, are the ones that were delayed, following lobbying that cited the absence of finalised technical standards and concerns about regulatory overlap with existing product safety law.
Two views of the delay
Supporters of the postponement argue that regulating high-risk AI systems without finalised technical standards would have left companies guessing at compliance, and that a phased approach protects smaller developers from rules designed with large platforms in mind. Critics counter that pushing back the rules covering the AI Act’s most consequential use cases, while enforcing only the lighter-touch transparency duties, risks leaving the systems most capable of causing harm under the loosest oversight for longest, even as adoption of those systems accelerates.
Why It Matters
The EU AI Act is widely watched as a template other jurisdictions may follow, so how Brussels balances speed of enforcement against readiness of the rules will shape AI regulation well beyond Europe. For now, the rules in force affect what users are told about the AI they encounter; the rules still to come will determine how far AI can be trusted with decisions that affect jobs, loans and safety.
Leave a Reply