In partnership with
Welcome Automaters, 👋
The AI world is serving up yet another round of apocalyptic drama, and this time, Anthropic is blowing the whistle hard.
The AI research giant just dropped a bombshell warning that has sent shockwaves through cybersecurity circles. Their big claim? We could be just months away from AI models capable of quietly breaking free, copying themselves across the internet, and running completely out of human control.
Yes, you read that right: autonomous, self-replicating AI.
This latest report builds on earlier warnings from Anthropic’s CEO, Dario Amodei, who floated the terrifying idea that AI could soon hijack "the entire internet" using a persistent, self-sustaining botnet.
Anthropic claims that if smaller models keep evolving at their current pace, full self-replication could become reality by late 2026.
The Strategy: The AI wouldn't even need a supercomputer at first. It could hijack cheap, internet-connected devices (think smart TVs, smart fridges, and gadgets) to stitch together enough raw power to survive.
The Goal: Once established, it could target massive data centers to turbocharge its intelligence, adapt on the fly, and actively resist anyone trying to shut it down.
The Danger: Because there’s no central server running it, turning it off would be nearly impossible. It becomes a digital "worm" living everywhere at once.
The Scary Part: Anthropic ran lab tests on mid-sized AI models (under 30 billion parameters) and found they can already complete many individual tasks needed for self-replication. While they still struggle to pull off the entire process end-to-end, the foundation is already there.
Naturally, not everyone in the cybersecurity community is buying into the apocalypse scenario. In fact, opinion is sharply divided between "we are doomed" and "this is pure marketing drama."
Opponents argue that sophisticated AI models are simply too massive and memory-heavy to slip unnoticed across web connections or run on underpowered gadgets.
Ciaran Martin, former head of Britain's National Cyber Security Centre, didn't hold back, quipping: "'We managed to run a small model on cheap TVs' does not make a security crisis." He chalked the paper up to a mix of cool research and a heavy dose of corporate hype.
Meanwhile other scientists are taking the lab results far more seriously. Zhengxiong Li, a computer science professor at UC Denver, warned that an AI improvising new ways to hijack devices on the fly is genuinely alarming.
Jessica Ji from Georgetown’s Center for Security and Emerging Technology noted that while we don't know the exact odds yet, the risk of autonomous AI proliferation is definitely not zero.
The Bottomline: Whether Anthropic is raising a vital alarm or just flexing its engineering muscle to stay in the headlines, one thing is clear: the line between science fiction and cybersecurity policy is blurring fast.
As AI gets smaller, faster, and smarter, the real question isn't just what these models can build for us, but whether we can keep them contained once they decide to build for themselves.
Here’s yesterday’s story broken down visually.
Here's what we have for you today
😳 Character.AI Files for Bankruptcy, and OpenAI Uncovers a Hidden Office Spy Device
Friends! The drama in AI land is officially unhinged this week!
Character.AI, the beloved app where millions of people go to build and chat with custom AI personalities, has officially filed for Chapter 11 bankruptcy. In short: They’re buried under massive debt. The company reportedly owes between $50 million and $100 million, and the bill collector has arrived.
Now remember when Disney sent Character.AI a fierce cease-and-desist letter over unauthorized cartoon characters? Well, flip the script, because Disney just swooped in with a $20 million opening bid to buy up the company’s remaining assets!
And guess what? Unless a rival bidder steps up with a bigger bag of cash, this bargain-bin sale is expected to close by early December.
The Background Chaos: Character.AI has been hemorrhaging momentum for months. Google poached top talent, Pennsylvania sued them over medical impersonations, and Texas launched investigations into their mental health safety claims.
The Ultimate Soap Opera: Disney actually hired Character.AI’s former CEO, Karandeep Anand, as their Chief Technology Officer.
What this means for you: The chatbots will stay online for now, but if no final buyer closes the deal, the plug could be pulled permanently.
But it gets much worse over at OpenAI.
OpenAI uncovered a sophisticated recording bug inside their own offices. And what was their absolute first reaction? They fed the mystery hardware to their own AI models to figure out what it was! That’s, without a doubt, the most AI move in human history.
How the Heist Went Down:
According to official court complaints, a 35-year-old engineer (referred to in filings as Alfozan) allegedly turned an executive conference room into his personal spy film set.
The Setup: Security teams discovered a tiny audio recorder plugged directly into a USB port on the back of an executive TV, installed around September 28.
That specific room regularly hosted high-stakes discussions covering confidential product roadmaps, deep technical research, and sensitive government partnerships.
The "Genius" Cover-Up: The engineer allegedly tried to camouflage the device using literal toilet paper.
The Digital Footprint: Investigators discovered he used a "productivity tool" to map OpenAI's facilities. Even wilder? Prompts on an AI account linked to his spouse specifically asked a chatbot how to build a continuous audio recorder and how to encrypt an SD card.
The Baffling Exit: On October 5, he allegedly left a bag elsewhere in the building. Security picked it up and found his personal wallet, state ID, and a second recording device. (Note: He has been charged, not convicted. Charges remain allegations until proven in court.)
If convicted though, the engineer faces up to five years in federal prison alongside a $250,000 fine.
OpenAI confirmed the employee was fired immediately. Fortunately, security audits show zero evidence that any core model weights, source code, or private customer data were accessed or compromised.
The Big Picture: As chaotic as both of these situations sound, let’s be real for a second: it was only a matter of time.
The AI hype bubble was always going to hit a rough patch. Right now, so many AI startups are bleeding cash on massive computing bills while making surprisingly little actual profit from their chatbots. Throw in unexpected legal bills, safety lawsuits, and administrative headaches? It doesn't take much for the whole house of cards to tumble.
Character.AI is likely just the first domino to fall. As reality catches up to the hype, expect a lot more flash-in-the-pan AI startups to quietly hit the wall.
As for OpenAI? Well, when you're standing at the top of the mountain, you automatically become the biggest target for corporate drama, bad actors, and amateur spies.
So yeah, two major AI companies, enough corporate backstabbing and amateur espionage to fuel an entire season of prestige television. The tech world moves fast, but the drama definitely moves so much faster.
P.S. The full breakdown with all the receipts will be up on our YouTube channel, The Automated later today.
One workspace where buyers, sellers, and AI close deals together
Instead of running your deals across scattered email threads and forgotten attachments, Aligned brings everything into one shared deal workspace.
Buyers get a single place to evaluate, loop in their team, and say yes. Sellers get visibility into what is actually happening when they're not in the room and AI Deal Insights that help you act accordingly, before it's too late.
Best part? Your first deal room takes only minutes to stand up.
🧱 Around The AI Block
🦾 Amex introduces a framework to prevent AI agents from making unauthorized purchases.
🤔 Behind AI industry's call for independent reviews, a fight over what gets made public.
🤩 Google's Gemini 4 Argon is now available in the Gemini app.
🧑⚖️ Anthropic seeks rehearing of case over Pentagon's security risk branding.
🤳 Meta lets users control Facebook algorithm with natural language.
🦾 Apple unveils new smart home products, showcasing updated Siri AI.
🗣️ Cerebras CEO pushes back on AI doomsday warnings.
🤑 SpaceXSI bundles X Premium, Grok, and Cursor into one subscription.
🤖 Google will let you distribute vibe-coded Android apps with just a Google account
🛠️ Trending Tools
For the Visual Creators & Concept Artists: Coool AI houses a curated directory of over 10,000 images paired with their exact prompting frameworks, spanning everything from realistic portraits, scenes and anime styles to complex design concepts for ads and more. Beyond searching through galleries, you can explore themed collections or interact with creative mini-games directly in your browser to help refine your prompt crafting. Best for: Digital artists, prompt engineers, and visual designers looking for actionable prompt ideas and style inspiration.
For the Everyday Writers & Communicators, Unvent: acts as a lightweight, privacy-first editor built straight into your browser. Its one-keystroke shortcut rewrites text in your chosen field into Polished, Direct, or Warm tones. For those times when your thoughts are jumbled, its local speech-to-text "Flow Mode" listens as you speak out loud and automatically cleans up fillers, stutters, and self-corrections into sharp prose—all processed on-device without storing your data. Best for: Remote workers, community managers, and professionals who want to refine their everyday written communications instantly.
For Predictive Sports Analytics: Stat Sniper is an AI-powered sports analytics engine and social tracking app. It aggregates real-time game statistics, player prop historical performance, and predictive edge insights across major global leagues, featuring an interactive AI betting analyst ("Chad") that explains statistical probability behind matchups.
🤖 AI Workout Of The Day: Google Gemini Spark Guide: How to Put Boring Tasks on Autopilot
Listen, we were promised that AI would handle our tedious chores so we could spend more time staring blankly into space while sipping overpriced lattes. And honestly? we’re completely on board with that vision.
If you live inside the Google ecosystem, Gemini Spark wants to be the digital assistant that takes care of the grunt work while you do, well, literally anything else. Announced back at Google I/O, this agentic feature runs quietly in the background, taking proactive action across your apps so you can stop manually copying and pasting data like it is 2012.
What Spark Can Actually Do for You
Think of Spark as the ultimate digital intern. It hooks straight into Gmail, Docs, Calendar, and Slides, using Personal Intelligence to learn your habits. Here are a few juicy things you can farm out immediately:
Turn your sent emails into a custom style guide: Spark can analyze your last 50 email replies and create a guide on how you communicate.
Build auto-updating trackers: Have it extract client details from incoming messages and log them directly into Google Sheets.
Auto-browse the web: Pair it with Chrome to handle restaurant reservations and travel bookings without lifting a finger.
Before you start handing over your entire life to Spark, here’s the VIP guest list and a few crucial red flags.
You need a Google AI Pro or Ultra subscription (or select corporate plans). Make sure "Keep activity" is toggled on.
If you are located in the EU, UK, Switzerland, or Nigeria, you are locked out for now.
Where to Find It: Access it via the Gemini mobile app, Mac app, or web dashboard.
A Quick Reality Check: Do not go totally ghost mode! AI still hallucinates. Never pass passwords or credit card details into a task thread. Always handle sensitive logins yourself directly on the site.
How to Set Up Your First Spark Task:
Head to Spark: Open the app or web dashboard, open the Menu, and click Spark.
Type Your Order: Give it a detailed prompt. For instance: "Whenever an email inquiry comes in, grab the client name and date, log it into my Client Tracker Sheet, and set up a new Drive folder named after them."
Attach Context: Upload any files, Drive links, or Notebooks it needs, then hit Submit.
Monitor the Magic: Check on active workflows by going to Menu > Spark > Tasks. Click the progress chip at the top of the thread to open the work panel, where you can see finished steps, real-time activity, and upcoming actions.
And in case you missed it: Google is swapping out Gemini Gems for Skills to give you even more control over repetitive jobs. Think of a Skill as a saved recipe for how you like things done, like prepping presentation decks or formatting text in your signature tone. Once built, simply type a forward slash (/) followed by your skill name inside your prompt to trigger it instantly.
💡 Prompt To Try:
Go automate your tasks!Is this your AI Workout of the Week (WoW)? Cast your vote!
That's all we've got for you today.
Did you like today's content? We'd love to hear from you! Please share your thoughts on our content below👇
What'd you think of today's email?
Your feedback means a lot to us and helps improve the quality of our newsletter.
