In the world, where LLMs are spitting out letter, words, code and contents faster that we humans can even think, I am becoming suspicious of every online content I consume. When, I am reading a blog, listening to podcast, I find myself suspecting if it is creation of humans ? I am guilty of using LLMs for my day-to-day task as well. For instance, writing an email, updating my resume and many more. But, for those, if I am on the receiving end I am not expecting senders to be creative, its just a mode of information transfer and if AI can make is succinct and easy to read, it’s good for me.

Same goes for code. There are so many coding agents, that can assist you or even own the code generation for you. Yes, on some level — as a developer I consider coding as an art and I’ve seen code from some of my peers that is just perfect to read and understand. But, just code in itself is not useful, what we care about is function that code can accomplish once it is executed. And if it functions what it says, I am good with it. So I don’t really care that much if that code is written by a AI agent.

But, when it comes to creative writing — for instance, a news article, a blog, a book, a research paper, yes I tend to care if it indeed is written by humans. Because, those are the mediums in which I invest my time, through which I learn new things, gets inspired to do (or not do) certain things, change my behavior, takes life decisions. And I want those contents to be authentic - coming from my fellow sapiens.

But, how do I know it’s authentic ? Unless someone send me a hand-written letter, it’s hard to get that assurance. What do we do for online content, which is the primary mode for me to consume them? Let’s keep aside two scenarios—humans can read from the AI and write them down char-by-char or we get robots who can mimic the human writing. I don’t have answers for those:(

But, for other common cases, can we get some attestation from the online sources about authenticity of the written contents ? We already have techniques that can works for different online editors like HackMD, wherein, the javascript captures the keyboard events from the client’s machines and detects if there are Copy+Paste, or bulk insert events. Even detect on the client, if some LLM agent was actively used during writing of this content. We can have content editors provider this as value-added service, wherein, it can attest the content creation methods used by authors and make it available as “verified by humans” tag on the content. It’s a win-win for authors and readers.

This simplified version certainly needs to be perfected to be used in practice, but, it’s just my personal food-for-thoughts. With some variations of such feature, whenever, I visit some online content, I can be assured to some degree about the authenticity of content creator and conent itself. Just like I can rely on Certificate Authority (CA) to attests the authenticity of the website I am visiting, a new Content Authority (CoA) can attest the authenticity of the contents on the web page. Well, in a feedback-loop, it might even be used by AI-bots crawling the web to prioritize the original content for model training or can be specifically protected from such bots.

“Attested - written by Human”