The EU AI Act officially entered into force on August 1, 2024, and its transparency protections for AI systems became effective on August 2, 2026. This two-year period, which has now concluded, saw European citizens exposed to AI systems operating without immediate transparency. The delay affects public interaction with AI, creating a period of uncertainty regarding its origins and intent.
The EU AI Act is legally active, and its core transparency protections for users were implemented on August 2, 2026. This tension means that while the law exists, its most significant user-facing safeguards remain dormant.
The immediate impact on user experience and AI system transparency will be minimal, potentially leading to a gap between public expectation and regulatory reality.
What Transparency Rules Are Coming for AI?
Specific transparency requirements are set to enhance user awareness of AI interactions. Certain AI-generated or manipulated content, including deepfakes, emotion recognition tools, and text published without human review, must be clearly labelled, according to the European Commission. Users must also be clearly informed when interacting with an AI system instead of a real person, as stated by the European Commission. These new transparency obligations aim to foster trust and integrity in the information ecosystem.
Why the EU AI Act's Transparency Rules Are Delayed
The two-year period between the AI Act's official entry into force and the implementation of its transparency rules meant that the very mechanisms designed to build trust were absent during the initial period of the Act's 'activity'. This delay could potentially erode public confidence rather than building it. The deferment specifically impacts protections against highly visible and potentially manipulative AI uses like deepfakes and undisclosed AI interactions, leaving a critical window open for misinformation and deception to proliferate unchecked under the 'active' law. This phased implementation created a paradox where the law was 'active' but its most direct safeguards against AI-driven deception were legally unenforceable until August 2, 2026.
