Brussels Effect reaches words AI creates.

Everyone’s familiar with AI and AI generated content. It isn’t a mystery how the latter has affected our lives. Fakes news, biased reporting, fake videos and social media clickbait has existed before. Only now, AI has amplified the scale, speed, cost, and also convincing nature of generated material. We feel it constantly. Even the youth isn’t protected from AI generated content. We went from Will smith eating spaghetti poorly to being akin to a movie. The speed at which this technology has evolved is astounding, and frankly, regulation and the part of the world has yet to catch up. Many people nowadays get confused and misinterpret AI as truth. So, what’s the answer?

As of the 2nd of August 2026, Article 50 of the European Union’s Artificial Intelligence Act started applying. It introduces new transparency requirements for providers and deployers of AI.

Requirements of Article 50 splits the responsibility between AI company/provider and person/company deploying the AI output. To make it short, the who, as in AI provider – Antropic, Google, OpenAI, will be required have their generated content as machine-readable marked and detectable. Chatbot providers will also be required to specify between interact from human or AI unless its obvious. Professionals publishing a deepfake will clearly have to disclose wether it was AI-generated/manipulated. This extends to Professionals who publish text on matter of public interests, unless there was human review/editorial control and someone assumes the editorial responsibility. To be specific, the EU does not require a ‘made by AI’ label on AI generated content.

The above highlighted comment refers to Article 50(4) and to quote: “-or where the AI-generated content has undergone a process of human review or editorial control and where a natural or legal person holds editorial responsibility for the publication of the content.” It poses an intriguing question, where does AI assistance end and where does AI authorship begins?

Anthropic states future Claude model will generate texts that contain a watermark. This way, it will determine the likelihood of determining whether Clause was involved in the writing of the text. Alongside Anthropic, others major AI providers are implementing this change to comply with the new requirements.

So if Claude writes 10% of something, what exactly should a detector say? What if human writes 90% Claude writes 10%, watermark might registered. What if Claude writes 90% and human rewrites enough, watermark might disappear.

Claude’s text watermark is a version of SynthID-text approach published by Google Deepmind published in a Nature paper 2024. For more information on SynthID, I recommend reading the article published by Google Deepmind. There they also mention the expansion of SynthID into video generated for Veo on VideoFX.

Watermark have both their advantages and disadvantages. They perform well under certain condition and are limited under others. Examples are they work well for large body of work. They lose reliability when texts are re-written, translated and modified.

WIRED reported developers have found ways to strip and bypass Anthropic’s watermark. Anthropic hasn’t yet released its official detection API, we don’t have definitive answers as of yet. Regardless if it works or not, it doesn’t render the EU law useless.

What we should be asking is this; does the transparency system have to be impossible to beat or good enough to retain the original provenance?

There is a need for regulation, technology advances fast, regulatory bodies must keep up. The inception of Artificial Intelligence was the same as opening Pandora’s box. It brought a lot of changes to the world and it brought it fast. Depending on where the user sits on the transparency spectrum, Article 50 will definitely affect how the world sees and consumes AI generated content going forward.

Thus by proxy of companies abiding by new regulations, Article 50 reached places such as Mauritius, the US, Japan, and many others where the EU law doesn’t apply. The law remains European, but its influence reaches far beyond Europe.

That is the Brussels Effect.