Anthropic plans to add an invisible mark to AI text-as the industry scrambles to police AI slop

Key Points:

• Anthropic is introducing an imperceptible, machine-readable watermark in its Claude AI-generated text and images starting August 2 to comply with the EU AI Act’s transparency rules, aiming to help identify AI-generated content without affecting quality or readability.
• The watermark is designed to persist through copying and some editing, signaling Claude’s involvement but not necessarily full authorship, and will be embedded across all Claude platforms including chatbot, API, and Claude Code.
• This move comes amid broader industry efforts to combat low-quality AI-generated content ("AI slop"), with platforms like Substack and YouTube implementing detection tools and policies to distinguish human from AI-generated work and regulate monetization.
• In related AI industry news, OpenAI executive Brad Lightcap announced his departure to start a new venture, House Democrats are pushing AI CEOs to testify on hacking incidents involving AI models, and Nvidia is partnering with major Wall Street firms to mobilize $500 billion for AI infrastructure financing.
• Researchers have used AI to design novel viruses capable of infecting bacteria, signaling both promising advances in phage therapy and raising biosecurity concerns, while Anthropic’s Claude AI made significant progress on the longstanding Riemann hypothesis math problem by increasing the proven share of a key set from 41.6% to 67.2%.