Hi there,

Writing this from a dock in Wisconsin. Short one, I promise. (well it’s short for me🫠)

Two things are getting lumped together right now in the AI world, and they’re not the same. If your eyes are already glazing on this line you might want to skip this week’s newsletter!

​Anthropic announced this week that AI-generated text is getting watermarked (and the other AI companies will be doing this shortly as well).

Say what does that mean?!

Claude is effectively baking a hidden signal into anything their model generates, to comply with EU law which is notoriously more strict around data and privacy.

Also side note I NEVER thought about this (I mean come on what is even privacy any more in this world… we’ve all sold it out for convenience) but truly, Europe cares way more about your privacy and data and I know this for fact on this trip back to the US.

But America cares more about you riding the lake in style….

But I digress…
​
Since I came home three weeks ago and switched to my USA SIM number, my phone rings LITERALLY every hour, sometimes more often, with absolute spam nonsense. Make this STOP!

My phone has never once rung in three years in Europe from any sort of spam or unsolicited call…so this is a small analogy but one that makes sense when done under the “to comply with EU law.”

If you want to run AI in Europe, you have to abide by their rules or else they will fine the hell out of you under a rule they passed in August.

And thankfully Europe is enough of a market where Claude is not just peacing out.

Ok back to what this means to be “watermarked”, especially if you are a Claude user…

Not much really.

Being watermarked is neutral. It just says “a machine produced these words.”
​
You could have the most creative, personal, brilliant exchange with Claude and the output would still get marked.

However there is another thing going on right now that is not good (in my opinion.)

And that is called “AI Detection”. **cue EVIL organ music

Detection tools -usually third-party ones – guess whether writing “feels” machine-made and then hand you a moral grade.

Not “a machine wrote this” watermark but more like “you cheated.”

Substack just implemented this a few weeks ago.

And I bet more and more of these “grader tools” get bought by mid-level managers who have no idea what they’re buying but have been told to “solve the AI problem” in their [insert industry here].

It FEELS productive.

But… ahem. I’m sorry, say what now?!

Let me get this straight.

The same AI these companies handed us -Claude and ChatGPT – was very much trained on human writing. (Just ask the authors in My StoryPro about the class action they’re bringing over their stolen IP.) ​
​
They built the machine out of us and now their detectors are turning around and flagging real human writing as “machine-made.”

So the detector is basically holding up a mirror and getting startled by its own reflection.

The original gets accused of copying the copy.
Make it make sense.

Someone in our My StoryPro community took her own podcast transcript, her LITERAL spoken words, out of her mouth, and ran it through the Substack detector.

And she was flagged.
Convicted of faking herself.

If your real words don’t count as real, I don’t know what the standard is anymore.

We are JUST starting to wade into these waters and this won’t go away as there is too much money in it.

The detection industry is already worth hundreds of millions and growing.

Turnitin alone charges institutions per-student, per-year and has since added an “AI-detector feature.”

Think restaurant letter grades in a NYC window.
Consumer Reports.
Whole industries are going to be built around certifying whether your writing is “clean.”

And who decides this you might be wondering…?!

Well PULL UP A CHAIR my friend!

The people building these tools decide.
Just like the people building these AI platforms get to influence so much of society right now.

So again, their world view, perspective, bias and all of that comes into the build and how they (mostly men…probably a lot of white ones) assess what is human and what isn’t.

So what do you actually do? (well you support AI built by minorities and women for starters…)
​​
You don’t try to beat the detector because you can’t win a game where sounding articulate is a crime. (EVEN if it’s “AI”.)

Vee told me the other day that if I was being “AI graded” right now in our meeting I’d be flagged at least 5 times in the first 5 minutes as even the way I spoke… which is how i’ve always spoken… would be considered an AI moral flag now.

Second, you don’t opt out of AI… that’s another newsletter because remember this is A SHORT ONE TODAY. #oops.

But you can go the other way.

You become so unmistakably, specifically, weirdly yourself that the question stops mattering.

This is the whole reason My StoryPro exists… not to just write for you but to help you identify all the weird, quirky, parts of yourself that you might not see.

Because the point is that when you build or write from your own life, your own voice, your own weird specific moments, there’s nothing to argue about.

It’s you…even if some detector would like to say it’s not so.

Your only defense against that is to keep showing up again and again and again being more of yourself.

Your quirky references.
Your actual opinions.
Your random Tuesday moment that somehow connects to what your audience is feeling.
The office hours story nobody else has.
The thing you said on a dock in Wisconsin on a hotdog float that made you stop and think. #happened.

That’s the premium skill now.
Being so consistently, provably yourself that no one has to ask.

The woman in our community with the flagged transcript…I more or less told her to “let it flag you.” Because anyone reading her words, her nuanced stories about her life knows that is from her and I’m sorry this is from her PODCAST?! Like it’s HER talking for real.

A detector can guess at patterns. It can flag a paragraph but it cannot fake a decade of you showing up.

The onus is on you to know how to do that. To tell the story only you can tell, in the voice only you have.

Become so recognizable that the question gets boring.

Which… not to be obvious about it… is exactly what we teach and how we built My StoryPro to do.
​
​Two things coming up:

AI Video Editing Workshop — August 27 (My StoryPro members only)

My StoryPro Relevancy Workshop — September 30

This one will be free for members but you can pay to be a part of this..
Signups will happen in the next week.

The whole premise is that research is commoditized but perspective is not.

We’re going to show you how to take any trend, any headline, any random personal moment and turn it into content that sounds like you.

The same skill I used as a TODAY Show producer to pitch the same story 15 different ways and get it approved every time.

That’s what we’re teaching.

More on September 30 dropping soon or you can sign up for My StoryPro and come to that one free tool.
​
Okay. Going to go lose to my kids at something or be freaked out again that this “aw take a cute pic of me and Cass pretend fishing” …

..turned into a LEGIT him catching a fish 5 minutes later and me freaking out that we were killing a fish.
​
Sound up if you need a laugh.

I was a vegetarian in another life for sure. Off to be a REAL person by the lake.

OK hope to see some of you at the workshops next month!! And has anyone else been flagged for being “AI” when they really wrote the thing? What are your thoughts on all of this?

XO
Patrice