1161 – Beware Of What You Can’t See: Tom talks Dangers Of Claude

SUMMARY BY CHATGPT

Tom Antion warns listeners about what he describes as a new practice involving Claude-generated text and invisible, machine-readable watermarks. He argues that creators should be cautious about relying heavily on AI-generated writing because automated detection could potentially associate their work with AI use.
Key points
• Claude as a writing tool: Tom says Claude is an excellent writing assistant and acknowledges using it—along with ChatGPT—to help create his updated 265-page automation e-book. He emphasizes that he substantially rewrites and personalizes AI-generated material.
• The main warning: Tom claims newer Claude models are embedding invisible, machine-readable information in generated text that may survive ordinary copying and pasting.
• Reputation concerns: His biggest concern is that simply detecting Claude's involvement could create questions about whether someone actually wrote their own material—even when AI was used only as an assistant.
• Proofreading concern: He suggests that even using Claude to proofread or make edits could potentially cause material to be identified as AI-assisted, although he acknowledges uncertainty about exactly when this would occur.
• Schools and plagiarism: Tom worries that AI detection could result in students being falsely accused of cheating or plagiarism, particularly if detection systems cannot reliably distinguish between AI assistance and completely AI-generated work.
• Control of detection: He is especially uncomfortable with the possibility that the company creating the watermarking technology could also control the tools used to detect it.
• Bad actors vs. legitimate creators: Tom predicts sophisticated spammers and scammers will eventually find ways around AI-detection systems, while ordinary writers and content creators could be the ones most affected.
Tom's recommended approach
His advice is to heavily personalize AI-assisted writing with genuine experiences, stories and knowledge that come directly from you. More importantly, he recommends being transparent about AI use rather than trying to conceal it—for example, acknowledging that Claude, ChatGPT or another AI system assisted with writing, research, editing or proofreading.
Rest of the episode
Tom promotes his free Automation e-book Version 4.0, his Great Internet Marketing Training mentoring program, and IMTCVA. He closes with criticism of traditional four-year college programs, arguing that career-focused technical education can sometimes provide more immediately marketable skills.
Key takeaway
Tom isn't recommending that people stop using Claude. His message is: use AI as an assistant, substantially make the work your own, and disclose AI assistance when appropriate rather than risking the appearance that you're trying to hide it.

===

Episode 1161 – Dangers Of Claude
[00:00:08] Welcome to Screw the Commute. The entrepreneurial podcast dedicated to getting you out of the car and into the money, with your host, lifelong entrepreneur and multimillionaire, Tom Antion.
[00:00:24] Hey everybody, it's Tom here with episode 1161 of Screw the Commute podcast. Today I'm going to give you a Claude warning. Ooh, scary. All right. Stick around for that. Hope you didn't miss episode 1160. That was How to Declutter Your iPhone. That was part of a three part series number 1150, 1155, and 1160 on taking care of all your pictures so you don't take 20 minutes to find one picture you want. All right. Pick up a copy of our automation e-book. Version 4.0 is finally out. Go to screwthecommute.com/automatefree to grab your free copy. It's 265 pages, but you don't have to read it all because each, just about each chapter has a practical section which you can just jump right to. That's version 4.0 and check out my mentor program. Greatinternetmarketingtraining.com. And my school, IMTCVA.org, certified to operate by SCHEV.
[00:01:29] Okey dokey. Yes, Claude, you got to worry about a little bit. Now I'm going to tell you ways around it and what to do and what's happening. Well, Claude is known as a really great writing machine. You know, this book I just told you about automate free was in part generated by Claude and then totally rewritten by me. But there's Claude is all over this thing along with ChatGPT and graphics. So yeah, I had enormous assistance with Claude to write this thing, and that's okay with me. I'm admitting it.
[00:02:13] And you can see it couldn't have possibly made up stories of me than stuff that I lived, so I customized it. But here's the thing. Claude is now adding invisible, machine readable watermarks to text generated by its newer models and. And that those are models generated after August 2nd, I believe is what what they said. And they're pushing it towards their old models. And it's happening on anybody that uses the Claude engine. So Google and Amazon I think use certain parts of it. So what's happening is, is you're getting this invisible machine readable stuff stuffed into your documents. Okay, now here's what With anthropic, the makers of Claude says about this, and I'll comment on some of these. So it says, okay, the reader can't see this. All right. That's nice. So your documents clean, copying and pasting probably will not remove the watermark. They claim there's no identifying info and that the these watermarks can't be traced to you. Well, to us this seems like just a stupid benefit because it's our document with our name on it. Or maybe in some giant company where 100 people work on a document, maybe that would apply. But for us, that's just stupid. It's our name on the document, so whatever's in it is traced to us. So that's that's a pretty stupid claim there. Now, these watermarks suggest that Claude was involved, but they don't tell if Claude wrote the whole thing.
[00:04:11] But here's the thing. If you're worried about this and your reputation, just having it come up at all in some detector or reader that's searching for these things hurt your reputation, so you're better off. This is my kind of fix for this is you're better off admitting it right up front, that the partial help with this document was through Claude. Now, they say that heavy rewriting or a complete rewrite can eliminate this. Another one. This is a perfect example for my upcoming show. Highly Educated idiots. All right. So. So if you were going to completely rewrite the thing or do a heavy rewriting of it, why did you bother using it in the first place? Right. So it's just stupid. These people have more money and Than brains and no common sense at all. Although they're just trying to see why a in this. Because people are are really upset about this example of this would be, let's say you had 50 different ways you could actually say a particular thing. Well, someone if you if you happen to have Claude say one of the ways. Well, someone with a decoder could recognize that text string and, and claim that Claude was involved in this. And like said, one claim could hurt your reputation. So you got this is something to really pay attention to. And, and here's the thing.
[00:06:01] Even if you just had it proofread the thing for you, it's possible this code will be generated. Now they claim if it just, you know, put a couple commas and periods or something like that. That that wasn't enough. That Claude would recognize text strings and and go ahead and put its watermark in. But how do you know? You can't be sure because there's no detect, you know, Anthropic controls the detector for this, which is one of the concerns about this. And so one of the things they were hit up about is why did you do this? Why? And one of their main arguments was European Union privacy concerns. So they're apparently strict over there on this this stuff. So that's what they claim. But also, you know, there was that guy on the radio last week that said that 32 of 35 students fell for his trick to see if they were putting the word Madagascar in their assignment for no apparent reason. Right. So you can look that up. So they're they're claiming what helps? Uh, find college cheaters. Uh, they also say it helps detect and, and so that you can eliminate total generated slop. The current term is fed slop for, for the narratives that they're trying to put out. But, but just in general information slop. So they're claiming that there's legitimate reasons. But on the other side of the coin, there's concerns that people could be falsely accused of plagiarism or, you know, not doing their assignments.
[00:07:53] There's privacy concerns. I mean, they claim, oh, we don't have any identifying information that could lead to you. But, you know, anytime something is tagged on your stuff, you know, who knows what could be done with it. Schools are worried about this in that someone could be called a cheater falsely. These are false accusations where they couldn't graduate or something, you know, because of something. One of these watermarks and the. The kid in school could have just had it proofread it for them, and then some detector shows up and says he cheated or she cheated. And then when one, you know, all the other detectors for AI generated stuff don't necessarily work on this. And, and so one of the concerns is, is only one company can control the detector for this. Hmm. Wonder what kind of can of worms that could happen when one company has the monopoly to say whether you're going to get recognized or not for, for doing AI slop, you know? So here's the thing. The super hackers will be able to defeat it. And only good people that are just regular authors and, and content creators won't know how to do it. And so the scammers will get away with murder on massive slop, probably with affiliate links and spamming and everything else and good people will get detected and and squashed.
[00:09:25] So that's another one that concerns so so my advice not so humble. I mean, I guess this is pretty humble advice because I don't consider myself some super hacker expert, but is customize whatever you write with stories that couldn't have been made up by AI with your specific things. Don't get a story out of some story speaker book and use it. You know, all the stories I tell are things that actually happen to me or somebody that I personally know that I know it happened. And so that's good. And then I admit it. Just say, hey, assistance and proofreading with this document were provided by Claude or ChatGPT, who or whatever you use. So that way you're not looking like you are hiding the fact that you use those tools. All right. So anyway, this is all kind of brand new stuff. Like I said, they just, uh, they're going to start the August 2nd was when it started rolling out on the new versions of Claude. So that was just a couple of weeks ago. All right. So that's my story and I'm sticking to it. Check out my mentor program, greatinternetmarketingtraining.com. And my school, IMTCVA.org certified to operate by SCHEV. What is SCHEV? SCHEV is the State Council on Higher Education in Virginia, which means I'm highly scrutinized compared to all these other, you know, rugrat wet behind ears, people that are doing this for two minutes and telling you how to get rich.
[00:10:57] So, so I got rich doing this stuff and still doing it because I love it. I could have quit years ago. So check out my school at IMTCVA.org. Probably save you a couple hundred grand in wasting it on a four year indoctrination camp, and you'll get an actual marketable skill that's in high demand. And in a lot of the the big companies, Google and IBM. And I think Nvidia and all the Intel don't require college degrees anymore. They're getting these people out of school. One of my assistants, a little boy who was I know when he was born and he just was here and he's in his second year of college and he's taken, which I mean, he's forced into taking so much crap like witchcraft history in the Middle Ages or something. And he's trying to be a civil engineer. Come on. All right. I know, uh, um, you know, it's good to have a rounded education, but you can't make a living on witchcraft. History in the Middle Ages. Let me tell you that unless you're really good at serving coffee at Starbucks. So anyway, check out my school. IMTCVA.org and I will catch you on the next episode. See you later.