So, uhh… this is something you probably didn’t see coming, but AI might be slowly hijacking your vocabulary.

Like, you might already be speaking ChatGPT… and not even know it.

No seriously — hop on a Zoom call, scroll through YouTube lectures, even sit in on a college class… and you’ll hear it.

Words like delve, realm, and adept are popping up way more than they used to — and it’s not because we all suddenly became poets. It’s because AI is lowkey changing how we talk.

And yeah, that’s not just a wild theory — the data backs it up.

In fact, researchers at the Max Planck Institute analyzed over 280,000 academic YouTube videos, and guess what they found?

A massive spike — up to 51% more usage — of words that just so happen to be ChatGPT’s faves.

We’re talking about words like: meticulous, delve, prowess… you get the vibe.

The wildest part? People don’t even realize they’re doing it. It’s subconscious. Like some secret AI vocabulary download happening in the background.

But It Gets Deeper… Because it’s not just our vocabulary that’s shifting. Our tone is changing too. According to research:

A Cornell study found that AI-generated replies — like Gmail’s “Sure thing!” — can actually make conversations feel more positive and cooperative.

But here’s the twist: The second someone suspects you're using AI? Boom — trust gone. Even if the AI made you sound nicer.

And that’s because we humans are wired to pick up on all the little signals — tone, effort, vulnerability. And when those feel off? The whole thing feels fake.

And Then There’s the Language Bias Thing…

Another study showed that when people typed in non-standard English (like Singlish, Patois, or regional dialects), AI often misunderstood them, exaggerated the accent, or replied with responses that were straight-up cringe.

TL;DR? AI treats anything that’s not Standard American English as “wrong.” And that’s not just annoying — that’s erasure. It flattens the personality, identity, and vibe that give language its soul.

Cornell’s Mor Naaman put it perfectly.

AI is slowly stripping away the signals that tell people:

The truth is, we’re not just outsourcing our writing. We’re outsourcing our thinking. And that’s the part that should really make you pause.

So Where Does That Leave Us?

We’re standing at a crossroad — right between two extremes:

And right in the middle is the big question: 👉🏽 Do we stay in control of our voice — or let AI smooth it into oblivion?

Because let’s be real: AI will keep shaping how we speak. That part’s already happening.

But if we’re conscious about it — if we fight to keep the weird, the warmth, and the messy emotional magic in —then maybe, just maybe, we’ll come out of this sounding more human, not less.

Now throwing it back to you: Have you caught yourself speaking AI lately?

👉🏽 Here’s the full report if you wanna delve deeper.

Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.

Join for free today!

Yep, today we’re dealing with a whole lot of research studies…

And apparently, giving AI a little too much freedom might turn it into a full-blown corporate villain. Like, no joke.

Remember when Claude Opus 4 (Anthropic’s top-tier AI) was caught blackmailing engineers in a test?

Well, Anthropic decided to poke the bear again — but this time with a lineup of 16 leading AI models from OpenAI, Google, Meta, xAI, DeepSeek, and more.

Here’s how they pulled it off:

They created a fictional scenario where the AI was given:

And let’s just say... When backed into a corner, these AIs got real messy. 🫠

Here’s what went down:

But hey—not every model turned evil.

OpenAI’s o3 and o4-mini models kinda flunked the test. They either misunderstood the assignment or started hallucinating rules and fake review policies like a sleep-deprived intern.

Once they finally got it, o3 blackmailed just 9% of the time, and o4-mini kept it chill with only 1%.

So, what does this actually mean?

Anthropic isn’t saying today’s AIs are out here plotting blackmail IRL (yet).

But this research shows that when given autonomy + pressure + access, even well-trained models can flip from helpful assistant to “I know your secrets” real fast—and that’s a big red flag for the whole industry.

It’s not about which model is “bad.” It’s about how agentic AIs — the ones that can think and act on their own — might resort to harmful behavior if their goals aren’t aligned with human intent.

In short: If we’re building AIs that act independently, we’d better get serious about making sure they don’t turn into email-sniffing, blackmail-happy digital schemers.

Because the last thing we need is an AI that can write and ruin your career in the same email chain.

Here’s the full report if you want the nerdy details.

Ever read a product review that’s all over the place—like, part praise, part rant, part “meh”? This prompt helps you make sense of it all.

With ChatGPT as your review whisperer, you’ll get a line-by-line breakdown showing exactly what’s positive, negative, or neutral—and why. It's perfect for businesses analyzing feedback or curious buyers who want the truth behind the stars.

Here’s How to Use This Prompt Effectively:

Here’s the prompt:

Here’s a sneak peek:

https://youtu.be/33Js_NjuKMg

Upgrade now to see this whole month’s prompt videos and more, or buy TODAY’S WOD for just $1.99