Anthropic's Decision and European Regulatory Requirements
According to a report by TechCrunch technology writer Lucas Ropek, Anthropic recently made a significant decision to add watermarks to the outputs of its artificial intelligence chatbot, Claude. This new policy involves embedding invisible code into the editorial text generated by the chatbot, which is designed to clearly mark it as AI-generated content.
Anthropic decided to implement this policy to comply with the requirements of the Transparency Code of the European Union Artificial Intelligence Act (EU AI Act). This law currently requires technology companies operating in the field to label and tag content that has been AI-generated or edited in a manner that allows computer systems to identify its algorithmic origin. While regulators in Europe may view this step as a positive and welcome development, various users of AI tools are expressing deep and prominent dissatisfaction with this decision.
The Reddit Storm and Theories Surrounding the Watermarks
The primary evidence of the brewing discontent among Claude users can be found on the internet platform Reddit, where a heated discussion has emerged, though not all users on the site agree with one another. One of the more dramatic and pathos-filled posts written on the subject was published by a user calling themselves visionode, whose Reddit account is relatively new, having been opened just three weeks before the comments were written. The user visionode claims in his post that the new watermarking system is a draconian conspiracy of sorts designed to victimize innocent chatbot users worldwide.
According to visionode’s line of reasoning, while experienced and sophisticated Claude users will be able to hide the evidence of their AI usage by paraphrasing the outputs or cleaning them through other AI services, the average, simple user of Claude will be the one who gets caught. According to him, these ordinary users, who use the tool for daily tasks, will be directly harmed by the company's move.
The Debate Over Ethics, Plagiarism, and Creativity
In his post, visionode demonstrated his claims regarding the users who would suffer from the new move: "Who will get caught? You. The student who used Claude to reorganize a paragraph. The journalist who asked the AI to summarize a two-hundred-page transcript. The writer who had creative block and asked for synonyms. Those guys come out of the process with a digital tattoo on their forehead."
TechCrunch writer Lucas Ropek notes in his article that without undermining the intensity of visionode’s outrage and frustration, the specific examples he chose to present are not the best or most convincing. A journalist who asks an AI system to summarize a long and complex two-hundred-page transcript should not be bothered or worried by a watermark attached to that summary, unless they choose to copy and paste that summary verbatim directly into their journalistic article—an act that is clearly considered unethical and should not be happening in the first place.
Similarly, the article notes that unethical behavior also occurs when a student copies and pastes Claude's outputs directly into their essay or academic paper after asking the chatbot merely to "reorganize a paragraph." Other Redditors were not quick to express support for visionode's outrage. One commenter wrote sarcastically, "Get a load of this guy," while another simply recommended that he "take a deep breath" and calm down.
The Question of Credit and Tool vs. Creator
The user visionode was not the only one complaining online about Anthropic's step. Another user, who was highly frustrated by the decision, defined the insertion of watermarks as "unethical" and "disgusting." The user argued that during their use of Claude, they themselves performed the lion's share of the hard and strenuous work. From their perspective, the chatbot was merely a "tool" designed to facilitate their arduous labor. The user explained: "I gave the instructions, context, decisions, and countless refinements, claude was the tool. If Claude starts watermarking the code or anything else it generates, what exactly is it claiming credit for?"
Many other Redditors also came out against these claims, dogpiling on the critic. "It’s not claiming credit though," one user shot back in the discussion. "It’s about being able to detect AI generated outputs because of the risks AI generated outputs can cause in various situations." Another user mocked the author of the complaint, noting derisively: "Bro couldn’t even complain about Claude without using Claude to write it."
Claims of Hypocrisy in Model Training
Some critics of the company's decision avoided presenting themselves as victims and tried to raise slightly more complex and nuanced arguments against Anthropic's new policy. For example, one user complained about what they called a general hypocrisy involved in applying watermarks to an editorial and textual product that itself was created and built by hoovering, collecting, and scraping the work and content of other people across the web to train the system. The user wrote: "I think it’s a very sinister direction to take. I don’t use Claude to write anything but having an AI that watermarks your work is terrifyingly ironic given how many of the frontier models came by their training data."
Most Users Support Transparency and Tracking
Despite the protesting and critical voices, the majority of AI users tended to express support for the new watermarking system, seeing it as a logical, reasonable, and necessary way to track content and materials generated by algorithms and computerized systems. A user on another thread summarized the issue decisively, writing: "There is literally no good argument for why this isn’t a good idea. The only reason you wouldn’t want this is to lie to people."