The European Union's Artificial Intelligence Act (AI Act) has transitioned from rule-making to practical application: as of August 2, 2026, new transparency and disclosure requirements for chatbots and AI-generated content have come into force in the EU. This was reported by Axios.
The European AI Office has not yet held any companies accountable for violations, but it has already gained the authority to request information from developers and demand access to models.
New Requirements Are Already Affecting AI Companies
One of the key innovations is the rules requiring user notification about interaction with AI and the use of artificial intelligence for content creation. According to Patrick Van Ecke, a partner at the law firm Cooley in Brussels who advises businesses on AI Act compliance, "it's a complex piece of legislation."
The expert noted that August 2 became the moment when ordinary users could begin to directly feel the law's impact—for example, by receiving notifications that they are viewing AI-generated content.
At the same time, companies are already adapting their products to the new rules. For instance, Anthropic announced that future Claude models will generate text with a watermark that is invisible to readers, while planning to provide users with an API for its detection.
Google and Meta also stated they are implementing and developing tools for transparency and labeling AI-generated content. Meta noted that it will work with the European AI Office to make the compliance approach "practical, interoperable, and genuinely useful."
OpenAI, for its part, is publishing summaries of the data used to train its models, as well as incorporating content provenance signals into them. Microsoft is adapting its internal governance and risk assessment processes in line with AI Act requirements.
The AI Act Could Become a Global Standard
Despite the complexity of implementation, the European rules could influence not only companies that directly operate in the EU market. Amy Worley, managing director and data protection director at Berkeley Research Group, believes the industry needs a single benchmark.
"We need a standard. This technology moment is moving very fast. And right now, at least, the EU AI Act gives us a standard," she stated.
According to her, requirements regarding notifications of AI use have already spread in the market, and watermarks could additionally be used by companies as protection in legal disputes.
At the same time, the August requirements are only part of the gradual implementation of the AI Act. Rules for key categories of high-risk systems, particularly in education, biometrics, and migration, as well as for AI integrated into physical products, will come into force later—in December 2027 and August 2028.
It is worth recalling that Pope Leo XIV recently called for strict regulation of artificial intelligence.
end-content
