🍜 A scammer used an AI MAGA..

Scammers have upgraded from fake princes to fake influencers...

Welcome, Noodle Networkers.

Meta may be watching everything, scammers have upgraded from fake princes to fake influencers, and one of the world’s most powerful AI models may have had some very unexpected visitors. Let’s get into it. Meta is reportedly considering tracking employee clicks, keystrokes, mouse movements, and even screenshots to train its next wave of AI. The company says it is about improving software, but to employees it probably feels like their computer just became a very judgmental roommate. Somewhere, someone is deleting their lunchtime solitaire history. 🖱️ Meanwhile, a scammer used an AI generated MAGA girl to trick gullible men online. At first she seemed charming, patriotic, and oddly interested in crypto. Then the victims realized the woman of their dreams was actually a laptop with a filter and a Venmo account. 😬 And unauthorized users reportedly accessed Anthropic’s Mythos model. This is the same cutting edge AI system that is supposed to stay very locked down. Instead, someone apparently found the back door and let themselves in. Nothing says “welcome to the future” quite like your billion dollar AI getting surprise guests. 🚨 From creepy office tracking to AI catfish schemes to mystery hackers poking around advanced models, the AI timeline just keeps getting stranger. Let’s dig in.

In today’s AI digest:

  • Meta may track employee clicks and keystrokes to train AI 🖱️

  • A scammer used an AI MAGA girl to trick gullible men 😬

  • Unauthorized users reportedly accessed Anthropic’s Mythos model 🚨

Read time: 5 minutes

WHAT’S HAPPENING TODAY

Sponsored by Somewhere.com

Hire top 1% global talent for 80% less than US equivalents. Join 4,500+ companies who have hired with Somewhere. There is zero risk to get started. You only pay if you hire.

(source: Reuters)

🖱️ The Digest: Meta may start tracking employee clicks, keystrokes, mouse movements, and even bits of screen activity to train its next generation of AI. Basically, your work laptop is no longer just a laptop. It is now a very nosy AI tutor taking notes like, “Interesting… Steve opens Slack 47 times before answering one email.”

Key Details:

⌨️ Your Cursor Is the Curriculum
Meta wants AI agents that can use software the same way humans do, so it plans to study how employees click around websites and apps. The future robot assistant may literally learn Excel by watching Karen spend 20 minutes trying to freeze the top row.

👀 No Hiding in Incognito Mode
Employees reportedly cannot opt out if they use a company laptop. Which means somewhere in Meta HQ, a product manager just realized their AI now knows they type “how to sound confident in meetings” into Google twice a week.

🛡️ Meta Promises It’s Fine
The company says the data is only for training AI, not for judging employees. So if you spend half your afternoon rearranging desktop icons or dramatically highlighting text while thinking, allegedly the bot is learning productivity… somehow.

🤖 Training the Robot That Replaces You
Meta says the goal is to build AI agents that can handle office tasks on their own. Which means employees are now in the very strange position of teaching a machine how to do their job better than they do. It is like being asked to train your replacement, except your replacement never asks for vacation days and somehow types 400 words per minute.

Why It Matters: This is a glimpse at what the future office might look like: AI that does not just answer questions, but actually knows how to use your computer. The awkward part is that before the bots can take over the boring tasks, they first have to watch us do them badly. Somewhere inside Meta, an AI is probably already learning that the first step in every corporate project is opening 12 tabs and panicking quietly.

AI scams

(source: Wired)

😬 The Digest: A scammer reportedly created an AI generated “MAGA girl” and used her to fool men online into buying fake photos and videos. The fake persona was presented as a conservative nurse named Emily Hart, which sounds less like a real person and more like the final boss of Facebook comments.

Key Details:

🤖 Meet “Emily Hart”
The scammer built an attractive pro Trump character aimed at older conservative men. Apparently catfishing was not enough anymore. Now it needs branding, audience targeting, and a campaign slogan.

💸 The Grift Was Weirdly Specific
After generic AI models failed to make money, the scammer switched to the MAGA angle and suddenly started cashing in. Turns out the internet’s most profitable niche might just be “lonely guys who think every profile picture is destiny.”

🧠 Even the Strategy Was AI Powered
The scammer says Google Gemini helped suggest the character and target audience. So not only did AI create the fake girl, it also became her campaign manager. Somewhere, a chatbot is accidentally winning Employee of the Month.

📱 The Men Fell for It
Victims reportedly believed “Emily” was real and paid for exclusive content. Which is a reminder that if someone online looks suspiciously perfect, they are either AI or a stock photo with a dream.

Why It Matters: This story shows how easy it is to use AI to create fake people that feel strangely real. We are entering an era where scammers do not even need to leave their basement anymore. They just need a laptop, a few prompts, and a fake profile named something like “PatriotPrincess1776.”

Anthropic

(source: Bloomberg)

🚨 The Digest: Anthropic’s secretive Mythos model was reportedly accessed by unauthorized users, which is exactly the kind of sentence that makes every AI safety researcher spit out their coffee. This was supposed to be one of the most tightly controlled AI rollouts in the world. Instead, it sounds like someone left the keys under the world’s most expensive keyboard.

Key Details:

🔓 The Wrong People Logged In
A small group of unauthorized users reportedly gained access to Mythos through Anthropic’s tightly restricted testing program. Anthropic has not said exactly what they saw or did, which is never the sentence you want to hear after the words “security incident.”

🧠 This Is Not Just Another Chatbot
Mythos is reportedly one of Anthropic’s most advanced models, built for high level cybersecurity and infrastructure tasks. This is not like sneaking into an unfinished AI image generator. It is more like accidentally wandering into the Pentagon and finding ChatGPT wearing a security badge.

🏢 The Guest List Was Supposed to Be Tiny
Anthropic had only planned to give Mythos to a handful of major companies and critical infrastructure groups. Somehow, the AI equivalent of an ultra exclusive rooftop party ended up with a few random people inside asking where the snacks were.

😬 The Timing Could Not Be Worse
The leak comes just as governments and regulators were already freaking out about Mythos and what it might be capable of. Anthropic spent months saying “don’t worry, we have this under control,” only for the universe to immediately reply, “that is adorable.”

Why It Matters: This is the kind of story that keeps AI executives awake at night and lawmakers reaching for the panic button. The whole promise of frontier AI is that the smartest models can be kept safe and controlled. If companies cannot lock down a test group, people are going to start wondering what happens when the next generation of AI is even more powerful. Right now, Anthropic looks less like the guardian of the future and more like the person who accidentally posted the Netflix password in the family group chat.

THE NOODLE LAB

Trending AI Tools

  • Snowpixel – AI tool that creates images, videos, and music from simple text prompts.

  • SlidesAI – Turns your text into presentation slides in seconds.

  • Remini – Enhances blurry or low quality photos with AI.

  • Quillbot Paraphraser – Rewrites text to make it clearer, smoother, or more concise.

  • Rose AI – Research tool that helps analyze data, documents, and reports faster.

AI Hacks & How-Tos

Rose AI is a research platform that helps you analyze data, documents and reports faster by bringing sourcing, cleaning, visualization and sharing into one workspace. Its documentation describes Rose as a “top down data tool” designed to make data analysis easy and transparent, and its platform includes notebooks, data search, analysis tools and sharing features.

How to Use It 🧭

1. Open a Workspace
Log into Rose and create a new workspace or notebook where your research will live. Rose’s workspace is where you create notebooks, search for data and run analyses.
Pro tip: Create one notebook per project so your sources and charts stay organized.

2. Find or Import Your Data
Use the search tools to find datasets inside Rose or upload your own files if needed. Rose is built to help users find, make and share important data in one place.
Pro tip: Start with one core dataset and add more only after your first analysis is clear.

3. Analyze the Data
Run your analysis inside the notebook and explore the numbers, tables or trends you want to understand. Rose is designed to support the full analysis flow inside the platform.
Pro tip: Focus on one question at a time such as growth, performance or comparison.

4. Visualize Key Findings
Turn your data into charts and interactive visuals. Rose documentation shows tools for line charts, timelines, multiple axes and data previews so you can inspect trends more easily.
Pro tip: Use charts for patterns and tables for detail so your findings are easier to explain.

5. Share the Results
Share datasets or notebooks with teammates using Rose’s built in sharing options and permissions.
Pro tip: Share only the final notebook view for cleaner collaboration.

Rose AI helps researchers and analysts move from raw information to clear insights faster, especially when working with data heavy reports and repeat analysis workflows.