Hello! 👋 🕵️ AI-written text may soon carry an invisible watermark, as per the the recent news from Anthropic, who are introducing invisible watermarks into text generated by Claude. You won’t see them. 👀 The writing will look exactly the same but there will be subtle patterns in the words Claude chooses that can make the text detectable as AI-generated. The technology is based on Google DeepMind’s SynthID system and is introduced secondary to new transparency requirements under the EU AI Act. Pretty sure that other major AI companies are working on similar approaches. 💡 Why this matters Well, we’ve spent the last few years asking: “Can you tell if this was written by AI?” That question may eventually become easier to answer and less subjective. Instead of teachers, employers or publishers trying to guess from someone’s writing style, AI-generated content could potentially carry its own invisible provenance. The key question to answer relates to accountability and ownership: if AI becomes embedded in Word, email, search, research and everyday work… when does something actually become “AI-generated”? If AI wrote the whole report? Easy. But what if it rewrote one paragraph? What if it suggested three sentences? What if you wrote it, AI edited it, and you rewrote the edit? 🤯 We may be getting much better at identifying where AI has been used at exactly the same time that the boundary between human work and AI work is disappearing. 👀 Should you care? yes… As AI becomes normal the question will shift from: ❌ “Did you use AI?” To: 🎯 “What responsibility do you still own for the finished work?” AI provenance can tell us something about how something was produced but it doesn’t answer whether that something is accurate, thoughtful, original or good. 💭 Question for you: If you wrote something, AI improved it, and you approved every word… who wrote it? 👇