Starting on 2 August 2026, the European Commission’s AI Office began implementing the long-awaited AI Act. This collaboration with national regulatory bodies transitions AI regulation from theoretical frameworks to real-world enforcement. New transparency mandates also went into effect on the same day. These require AI systems to communicate clearly when users interact with automated technologies. They also require users to know when content is created or modified by artificial means. The requirements aim to give people full visibility into AI interactions. They also ensure visibility into the digital content people engage with.
Transparency for Users and Businesses
The AI Act mandates that chatbots, deepfakes, and all forms of AI-generated or edited content must include explicit disclosures. For example, chatbots must clearly identify themselves as artificial systems, and deepfakes must be labeled as such to prevent misinterpretation. These measures aim to reduce the potential for deception and enable users to make educated choices about the authenticity and purpose of the content they encounter.
Machine-created or edited content must include machine-readable tags. These tags let platforms and users detect the origin through automated tools. This offers a practical way for users to trace information sources. It also helps businesses demonstrate that their AI systems are transparent. The transparency requirements, as outlined in the Act, provide a clear compliance framework for businesses. They allow companies to continue developing and deploying AI without unnecessary roadblocks.
Who’s in the Spotlight
The new enforcement powers under the AI Act primarily target providers of general-purpose AI (GPAI) models. These models are the underlying engines for a wide range of AI tools and services, including AI agents that automate tasks across industries. Because these models are so widely used, they are a central focus for the Commission’s regulatory efforts. The Act requires providers to address potential high-risk scenarios, such as threats involving chemical or biological agents, cyber attacks, and breaches of fundamental rights.
Under these requirements, providers must not only document specific information but also provide it to relevant authorities or downstream users. This includes developing a copyright policy and publishing detailed summaries of the training data used for their models. The Commission states that this information should give stakeholders a clear view into how the AI systems operate and what data they rely on to generate outcomes.
Failure to comply with these obligations could lead to substantial penalties. Fines can be up to €15 million or 3% of a company’s global annual revenue. The higher amount applies in either case. These financial consequences place major tech companies under intense regulatory pressure. OpenAI, Anthropic, and Google are among those affected. They already face various digital regulations across the EU. Henna Virkkunen, Executive Vice-President for Tech Sovereignty, Security and Democracy, described AI as a powerful tool. She said it has the potential to benefit both people and businesses. She warned that the most advanced models could lead to unprecedented risks.
The enforcement of these rules might intensify tensions between the European Commission and U.S.-based tech giants. The Commission is pushing its agenda of tech sovereignty and data protection. The EU seeks to align AI development with public safety and consumer rights. It also aims to support innovation in the field.
Not all industry experts agree that these regulatory efforts will produce meaningful change. Some argue that the effectiveness of the AI Act might not depend solely on legal requirements but rather on increasing public awareness of how AI processes and handles data. Antonija Vojnović, Governance, Risk and Compliance Department Manager at Span, suggested in a May 2026 interview with Help Net Security that user understanding and awareness of AI’s role in data processing could be just as important, if not more, than the laws themselves.