Why Watermark?

Speaker

[music]

Andrew Keen

Hello, everybody. It's Saturday, August the 15th, 2026. I'll probably put this show up tomorrow on the 16th. The heart of the summer. Everyone's away. I'm not quite sure why we're doing these shows. And according to the Wall Street Journal, this is the summer that America became a nation of Luddites. And it's not just the left that's turning against technology. It's also the right. There was an interesting piece in Transformer. About a Jan-6 organizer who's getting conservatives to be riled up about AI. Everyone seems riled up about AI. And according to Keith Teare, the publisher of That Was The Week newsletter, it's not just the left and the right, it's even the AI community. And his AI, not his AI, that was an error, in his editorial this week, he asked why Watermark? Suggests that Claude, in supporting or introducing watermarking in its product,

Andrew Keen

is suggesting that my product is so bad for mankind that I plan to place an indelible fingerprint that I created it so everybody can know I was there. Keith, before we get To your views on watermarking, you're not a fan, you're not keen on watermarking, to borrow some of my language. What exactly is watermarking?

Keith Teare

Well, it's a word that describes lots of different techniques for identifying the source of content. For example, Google for a long time has used audio signals to watermark videos for copyright holders. In this case, it's anthropic using text signals. In other words, it's adding stuff invisibly to your writing where it can say, yes, that came from us.

Andrew Keen

So it's just, it's indicating whether or not you or I or anyone else is using AI. And you... Asking your editorial this week, why Watermark? Claude wants everybody to know it is there, which actually isn't surprising because Claude wants everybody to know it's everywhere. But what's wrong with watermarking? Why are you against it?

Keith Teare

I'm against it for a lot of reasons. And I will say to caveat, I don't know that I'm against it per se. I'm against it in this context.

Andrew Keen

Well, you say, and I'm quoting you in your editor, you say, watermarking is nuts. So you couldn't say it more clearly than that, unless you like nuts, Keith. Do you like nuts?

Keith Teare

Just had them for breakfast, almonds. Look, watermarking this week is nuts. Let's have this week, because it's really about anthropic. For a reason I can't quite fathom, Deciding that it will add something to the world if it watermarks things it's involved in. And in my brain, the only rationale I could come up with for that is they're almost like ashamed to be there. And they also don't want their customers to use them without whoever is reading their customers to know they're there. So it's kind of tarnishing Customers' use of their product by implying there's something wrong with using it. And so it's a quite remarkable decision that they've made to do that. And it comes this week with lots of whispers, which are not so quiet, that internally Anthropic is turning into a little bit of an ideologically motivated cult. And that word cult

Andrew Keen

is being used. Cult rather than cot. I'm thinking of a baby, a cult. Well, you hate the anthropic. Let's be clear, Keith. You've never barely said a good word about it. You're an open AI guy.

Keith Teare

Well, this week I think I've become a grokbot guy. I'm for sale, actually. You are? I don't really owe any long-term affinity to any of them. And I've always said Anthropix Claude is excellent, and it is. And I think OpenAI's codex-ChatGPT is excellent. This week, You have to say Grockbot is excellent. It's as good as the others.

Andrew Keen

I mean, this is our wacky race is every week things change. Although the FT today leads with a piece about OpenAI upheaval mounts as Sam Altman read his IPO push. I mean, I don't know how it can mount much more. Every day there seems to be a headline in the FT or the New York Times or the Wall Street Journal about upheaval at OpenAI. And the same is true of Anthropic in terms of their supposed $2 trillion

Andrew Keen

Keith, let's get a little bit more into this idea of shame or guilt. Why in your mind is watermarking introducing the issue of shame or guilt to the use of AI? Why does it, in your view, and you suggested at least for this week, AI, watermarking is nuts, you're not a big fan of it. Why does it make its users feel guilty?

Keith Teare

Well, that paragraph you just showed explains it. A watermark starts from suspicion. Why? In other words, there's a problem to solve. The watermark is solving some problem. Well, what is that problem that it's solving? I don't see a problem with using AI, so I definitely don't think there's a problem. But there is a zeitgeist. I don't know if it's the zeitgeist. That use of AI somehow tarnishes the user. And it feels as if Anthromic is embracing that criticism and sees watermarking as a solution, which, if you think about it, if it has any effect at all, people on the margin will therefore not use it.

Andrew Keen

But I don't understand... I wonder whether there's a kind of guilt here with you. Why would people be embarrassed to use anthropic? I'm not embarrassed to use it. You're not embarrassed to use it. We both acknowledge it. Why would watermarking it suggest guilt? Why wouldn't people wear it as a badge of pride?

Keith Teare

I think we've got to ask Anthropic that question.

Andrew Keen

No, but it's your analysis. It's not Anthropic. Anthropic's just providing the watermark, and they're not suggesting that using it is a bad thing. That's your interpretation.

Keith Teare

Well, why else would you need a watermark? Try to give me three other reasons why you would use a watermark.

Andrew Keen

Well, let's just say, I'm curious. So one of the other pieces you linked to this week is one by Alex Danko about And it's a headline, this essay is 10% AI generated. I'm always curious about whether or not people are using AI. It doesn't mean that it's a bad thing. And there are now products where you can run content through, text through to determine. So I ran your editorial this week through Pangram, and it suggested that 86% of your editorial is

Andrew Keen

Is produced by AI. I actually ran last week's editorial of mine. I couldn't do it this week because I haven't written it yet. And it said that mine is 100% human written. But that doesn't make me feel any better or worse. It shouldn't make you feel any better or worse. The fact that you're producing 86% of your content with AI doesn't make you feel inferior to me, does it, Keith?

Keith Teare

No, but it's also wrong. I put Saul Klein's story this week through it, and it said it was 100% written by AI. Well, we don't know. We'll have to get Klein. Maybe it is. And I read his piece, and I could definitely see bits of it that I thought AI was involved in, but it wasn't 100%. And even if it was 100%, I could identify what I know about Saul's views completely penetrating that piece. So, you know.

Andrew Keen

But again, you're the one, Keith, it sounds to me, I'm not a psychologist here, but you're the one.

Keith Teare

You're not being fair. You're ignoring evidence. Last week we talked about publishers and a writer having his book cancelled and his agent resigning because he used AI. Yeah, but we're not producing. Don't pretend there isn't an issue. There is an issue. It's not me.

Andrew Keen

But the issue then is a social one that for some reason people feel that you shouldn't be using AI. But that's got nothing to do with Anthropic.

Keith Teare

No, Anthropic's responding to that feeling.

Andrew Keen

Right. So that doesn't mean that it's suggesting that. But the interesting thing is why people are feeling it.

Keith Teare

You shouldn't respond to that feeling because it's wrong. If I'm leading Anthropic, I call bullshit on the publisher. I do what Elon would do. I don't do what Dario would do, which is kiss ass. And by the way, he kisses ass completely in kind of a deceitful way where he's just trying to adjust to what people are asking. He reminds me of a politician, not the leader of a company.

Andrew Keen

But don't, I mean, in defense of Dario... Probably doesn't need much of a defense from me. If indeed America is becoming a nation of Luddites, which the Wall Street Journal announced today, for better or worse, it's not saying that's a good or a bad thing. Both left and right are more and more suspicious of AI. If you were running Anthropic, don't you have to adjust to that? Isn't that just the reality of the market?

Keith Teare

You should never adjust to the market. The market's wrong nine times out of ten. You have to adjust to the long-term roadmap, the capabilities, and the good things it will bring to humanity if you succeed. You have to fund that, you have to earn revenue through that, and you have to either ignore or call out the people who are friction in the way of your roadmap. That's the right way for any leader of any startup to behave, because a startup that's disruptive inherently We'll have opponents, usually people that don't want change.

Andrew Keen

And I'm trying to get out, whatever, whichever position they take, they're going to have opponents. That's just the nature of things. And I wonder, you always use the H word, humanity or human. I never know quite what it means, but it seems as if this issue, Keith, of whether or not we use AI and the shame that it involves, I sense a little bit of shame on your part. Maybe I'm wrong. Maybe I'm the wrong kind of therapist here. That's all too human. And that's a good thing, isn't it? I mean, none of us want to become machines. We know that the future is of increasingly smart machines. We're trying to carve out a space, a place where humans will remain human. And isn't shame one of those spaces? Because then it's not really rational, as you say. I mean, but it's all too human.

Keith Teare

I think shame should exist. And I would feel shame. If I was replacing myself with AI and not caring what it said, I'm publishing it anyway. If it really was, and there are spam factories using AI that do exactly that. There are also fakers who do exactly that. And if I was doing that, I definitely would feel shame. I don't feel any shame at all recruiting an army of AI agents to help me accomplish my goals. None whatsoever. As long as I can stand by the output, which I therefore have to read it, change it, edit it, inspire it in the first case, then I don't feel any shame at all. But I think society is trying to make me feel shame. And as a normal human being who listens to everything, you certainly, and this must be true of you as well, Andrew, you must ask yourself, you know, always, have I crossed the line where those people are going to Was it right, by the way, was it right

Andrew Keen

to give you 100%? No, because I use Anthropic, but I use it

Andrew Keen

As a draft, and then I rewrite the whole thing.

Keith Teare

So it gives us some structure, which is exactly what I do. Actually, there's a pre, you know, you're cutting yourself short there, you give it some content.

Andrew Keen

So you don't Yeah, well, it has the it has the interview. So we create a transcript, then I say suggest I suggest some notes, and then I mostly rewrite it, but I use sometimes its structure and some of its ideas.

Keith Teare

So I do exactly the same. I give it content. In my case, it's the newsletter. In your case, it's the interview. So it starts with your work, my work. Then...

Keith Teare

I'll ask it, how many themes are there this week in the stories? Usually on a Friday, there's 50-ish stories. It will surface four or five themes that are buried inside the stories with evidence that the themes are there, using quotes and what have you. And I'll say, a bit like a news editorial meeting, I'll say, actually, I like theme one, but it could be weaved in Logically, throughout the whole thing, there could be a line where all the themes come into play, but it's through the spectrum of theme one. Go and do me a draft editorial that does that, and it will do it. It'll be about 90% not how I think. Then I'll go back and forth a few times, tweak it. I get a draft at the end that's half decent, and then I rewrite it. That's what I do.

Andrew Keen

But again, this kind of conversation is... An all too human conversation. We are, we're speaking of how we each produce our unique content. I don't see why watermarking, I think watermarking in a sense is a very human addition To a product that in some people's minds is certainly in America this summer is threatening our humanities. I don't have a problem with it at all.

Keith Teare

I think it's a good thing. Why isn't Restream watermarking this saying it came from Restream? Why doesn't Adobe Premiere? Well,

Andrew Keen

if it did.

Keith Teare

Yeah, but it's bullshit. There's no need for it. It's just a defensive reaction to a zeitgeist.

Andrew Keen

Well, you're not alone. I disagree with you, but lots of people are angry. One of your posts of the week is from someone called Arturo Villa. I wonder if it's related to Aston Villa. Anyway, he just cancelled his Claude Mac subscription. So you're not alone. Did you cancel, Keith? Maybe if you felt strongly enough, you should just cancel. Do it. Go public with your outrage against Claude for watermarking.

Keith Teare

Well, this probably gets to the nuance of what I'm saying. No, I haven't cancelled it. And honestly, I don't care that it's watermarked. I care that Anthropic thinks it should be watermarked, but I don't actually care that my stuff is watermarked. Why do you care what Anthropic thinks? Because I'm an analyst of tech, and I think it's a huge mistake. One of several is made, or they've made. And, you know, I think they're in danger of destroying their company in the eyes of those who most use their product.

Andrew Keen

Yeah, I mean, my problem, actually, I'm a big fan of both Dario and certainly compared with some of the other characters like Elon and Sam Altman. My problem with Anthropic this week, I didn't tell you about this, is that I'm already on a $100 monthly subscription, which seems a lot to me. And my sense is that because they're trying to drive more and more revenue for the upcoming IPO, a supposedly $2 trillion IPO, they're narrowing down on use. So I used it once, and it's now telling me that I couldn't use it. So I wonder whether that's going on. And again, I'm sure that Did you run out