Tag: watermarks

Recently, Anthropic decided to implement a new policy of watermarking Claude’s outputs by embedding invisible code into the chatbot’s editorial text to indicate that it was AI-generated. This decision was made in order to comply with the EU AI Act’s Transparency Code, which now mandates tech companies to label content that has been AI-generated or edited in a way recognizable to computer systems. While European regulators are pleased with this move, some AI users are not in favor of it. Evidence of discontent can be found on Reddit, where users like visionode have expressed strong opposition to the new watermarking system. Visionode believes that the system is a draconian conspiracy meant to target innocent chatbot users worldwide, arguing that it unfairly exposes users who rely on Claude for various tasks to a digital tattoo of sorts. However, not all Reddit users share visionode’s perspective. Some have pointed out that the examples used to illustrate the issue are not the best, as ethical concerns arise when users copy Claude's outputs verbatim into their work. Others have supported the watermarking system as a necessary measure to detect AI-generated content and mitigate potential risks associated with it. Critics like visionode and others have raised concerns about the ethics of watermarking content produced by AI, especially considering that the AI itself has been trained on existing data gathered from various sources. While some users view this as a sinister development, others see it as a sensible way to ensure transparency and accountability in the use of AI technology. Overall, the debate around Anthropic’s new policy continues on Reddit, with users expressing a range of opinions on the issue. Some see it as a necessary step to prevent deception, while others question the implications of marking AI-generated content in this manner.

Tag: watermarks

Some Claude users are mad that Anthropic’s new watermarks will catch them using it at their jobs, classes

Quick Links

Resources

Support