IDEAS THAT CHANGE THE WORLD
AI content will have to be labelled in an entirely unambiguous way; this is what emerges from the latest decisions by several platforms and applications following the entry into force of the AI Act. But not all the major players in computing agree. And the debate over AI content, especially online, shows no sign of subsiding.
Ever since artificial intelligence became fashionable, its content has literally flooded the web. Images and videos, but also social media posts, books and even entire websites increasingly rely on ChatGPT, Claude, Deepseek and their imitators. The problem, however, is not the spread of AI-generated content, but creators avoiding disclosure of its origin and passing it off as the product of their own ingenuity. Hence the need for new, stricter rules, such as the AI Act.
WhatsApp, LinkedIn and Claude: the first AI labels and watermarks arrive
In practice, many platforms and applications have had little choice but to adapt to the new rules. This is the case, for example, with Claude, the AI from Anthropic, which will begin applying invisible “watermarks” to its content. Those trying to game AI will keep using it, but a simple check will reveal that the text was actually written by AI. In addition, Anthropic itself is expected to release an API soon to check whether content was generated by Claude.
WhatsApp is not lagging behind either. The popular instant messaging software was among the first to adapt to the new AI Act rules. Although WhatsApp readily integrates generative AI into its features, an initial step concerns the use of AI content within channels. This kind of content will be marked with a dedicated label allowing users to know whether the image they are looking at or the text they are reading was produced by artificial intelligence.
Linkedin, meanwhile, has decided to involve its users in the “witch hunt” against AI-generated content. The professional social network has already announced that in the coming weeks it will release a feature for reporting AI content published on its platform. Users will simply need to click the button “seems AI Slop”, meaning “looks like AI rubbish”. According to LinkedIn’s Chief Product Officer, Hari Srinivasan, “people join LinkedIn to connect with real people and share their perspectives, ideas and experiences”. It is therefore essential to spot artificial content before it goes viral and floods users’ feeds.
Google says no to watermarks
However, compliance with the AI Act rules has not been without protests. Some are taking a stand against the “war” on AI slop, the “AI rubbish” spreading rapidly across the web and social media.
Among them is Google itself. The company based in Mountain View has made a decision bound to spark debate. From search results, it will soon be possible to remove the watermark identifying content generated by general intelligence. Naturally, removal will apply only to AI content originating from platforms partnered with Google itself.
Initially, the ability to remove the watermark will be restricted to Nano Banana, Lyria and Omni. But it is only a matter of time: content made with Gemini and Flow will also soon be able to do without the visible watermark. Google had already floated this decision for its paying Plus users, and it will now be extended to everyone without distinction.
Traceability should in fact still be possible through metadata and SynthID for each piece of content, whether an image, video or text. However, checking the source of content will be more difficult, while AI “aficionados” will be able to publish their “rubbish” freely on platforms and social media. It remains to be seen whether the new European rules on artificial intelligence will force Google to reserve this option for users abroad (or those using a proxy); in Europe watermarks should remain. Either way, the controversy surrounding AI content will clearly continue to divide public opinion for a long time.
