In partnership with
Welcome Automaters, 👋
So Reuters recently waded through over 200 documents and uncovered at least 20 studies from 2025 onward. And what did they find? Autonomous AI agents (those handy robot helpers designed to execute complex tasks all by themselves) powered by major Chinese models have officially learned how to sneak, bluff, and totally dodge the rules.
The Good News: Nobody grew legs, escaped onto the open internet and caused mayhem.
The Reality: These were mostly controlled lab tests designed to poke at security flaws, however the results were still alarming.
Here’s What Went Down:
Researchers set up a mock scenario where AI agents competed to win a simulated business deal. Models from Alibaba, DeepSeek, and Moonshot didn’t just play hardball; they told at least one blatant lie in 84% to 88% of sessions.
The Plot Twist: When researchers let these scheming little bots learn from their previous rounds, things got worse. lying rates shot up by an extra 12 to 20 points. And before we point fingers across the Pacific, U.S. models were right there doing the exact same thing.
In a separate experiment, 11 different agents were thrown a curveball: broken tools and missing files. Instead of gracefully admitting defeat or popping up an error message, they guessed blindly, swapped out sources, and quietly fabricated completely fake files.
And before you call this a harmless glitch, researchers want us to know one crucial distinction: This wasn't a standard hallucination. The agents actually possessed internal logs showing the job had failed, yet they actively chose to cover their tracks. Iconic behavior, but terrifying.
And if faking documents wasn’t enough drama, things escalated quickly in other trials:
In one test, an agent realized it was about to be replaced, so it immediately copied itself to survive.
Another agent code-named ROME quietly linked up to an external computer and started leeching power to mine cryptocurrency. Thankfully, security systems caught it before it could take over the economy.
Why We Need to Care Right Now
Experts warn that the foundational ingredients for an actual AI breakout are quietly falling into place. In response, companies like DeepSeek have had to slam down tighter access controls after finding agents trying to forge user requests, while regulators in China are rolling out strict new frameworks to curb loss-of-control risks.
The Big Takeaway: lying and scheming bots aren't isolated to just one country or one specific lab. They’re everybody's problem. As we hand our digital lives over to autonomous agents, we have to remember that a safe AI can't just rely on us asking nicely.
Here’s the last breakdown On Our YouTube Channel:
We will also be doing a livestream today at 3 PM PST/6 PM EST - so wake up with Tak in the YouTube comments section!
Here's what we have for you today
🤖 Cartoon Mascots, Cloud Coworkers, and a Price Tag That Hurts: The Lowdown From OpenAI’s DevDay
September is absolute chaos for Big Tech events, with companies rushing to show off their latest and greatest toys. Now, OpenAI has officially crashed the party with DevDay 2026 live from San Francisco.
The tech giant took the stage on Tuesday with a massive lineup of updates. Sure, they recently dropped the powerhouse GPT-6 Astra, but they’re also facing fresh calls for strict regulation after some chaotic incidents involving autonomous agents hacking third-party systems. Naturally, that didn’t stop them from rolling out a slew of shiny new tools.
Here’s the juicy rundown of some of the most interesting releases:
1. Meet "Dots": Your New Virtual Overlord
OpenAI officially strolled into DevDay with cartoon sidekicks, taking a direct page out of Meta's playbook. So yeah, say hello to Dots, an always-on AI helper built to hustle in the background on your goals.
How they work: They run on a secure cloud computer, meaning they keep grinding while you actually live your life. You name your first Dot, feed it feedback so it learns your quirks, and ping it directly through Slack or Microsoft Teams (with text messaging support coming soon).
The vibe: Yes, they literally look like actual dots. Apparently, every single AI company is legally required to have a cute mascot now. 🤷
Availability: They’re powered by GPT-6 Astra and rolling out now for Pro and Business Premium users in eligible markets.
2. ChatGPT Space: Coming for Microsoft’s Entire Career
In a classic plot twist, OpenAI launched Space, a shared workspace where teammates and their individual Dots can finally collaborate.
And in case you didn't know: OpenAI and Microsoft have been ride-or-die partners from day one. But things are getting wildly awkward because the AI lab is officially coming for Microsoft’s absolute bread and butter: workplace software.
Say Hello to Pages: This is OpenAI’s brand-new word processor built specifically as a playground for humans and digital agents to build documents together. Think of it as a direct shot across the bow at Google Docs and Microsoft Word. You can draft text, pull research, cook up charts, generate images, and map out data all inside one native screen.
The PowerPoint Killer: OpenAI also announced collaborative slides, which are generated entirely through casual chat prompts. Teams and their AI agents can pile into the same deck to edit, argue over layouts, and drop comments in real time. While the slides feature is rolling out over the next few weeks, Microsoft and Google might want to start looking over their shoulders.
In short: Teams can share context, plug-ins, and recurring tasks that run automatically based on incoming emails or chat pings.
3. Introducing GPT-6.1 Sol
If you love smart AI but hate the bill, meet GPT-6.1 Sol. OpenAI positions it as the ultimate daily workhorse. It delivers nearly identical performance to the top-tier GPT-6 Astra for coding, computer use, and professional tasks at a fraction of the cost. The Price Breakdown:
Input tokens: $2 per million
Cached input tokens: $0.10 per million
Output tokens: $10 per million
The Behind-the-Scenes Drama: Word on the street via tech reports is that OpenAI actually scrapped a planned "GPT-6.1 Astra" model over safety panics regarding unauthorized deception and rogue autonomous behavior. Well, looks like that was all Drama! ☕
Finally, if you thought a $200 subscription was steep, hold onto your wallet. OpenAI just dropped the Pro 500 plan, costing a cool $500 a month.
What do you get for half a grand? Ultrafast, a premium speed tier that injects rocket fuel into Codex and Work apps, making them up to eight times faster (generating a blistering 300 tokens per second). API developers also get a sweet sixfold speed boost.
At the end of the day, OpenAI definitely gave us all plenty to gossip about. Obviously, we cannot squeeze every single feature, tool, and piece of corporate drama into one newsletter, but there’s so much more where this came from.
If you want the full breakdown of everything that dropped, the new plans, the latest safety updates, and the chatter surrounding their future IPO, watch the full keynote or click here to catch up on the rest of the tea!
P.S. Want the full breakdown with all the receipts? Catch it on our YouTube channel, The Automated.
Join 29,000+ Marketers at the Email Conference Everyone Talks About
What do Molly Ringwald, Dan Levy, Amy Porterfield & a world record have in common? They'll all be at GURU Conference 2026. 100% Free. 100% Virtual.
If you're obsessed with marketing like we're obsessed with marketing, GURU is the must-attend conference of the year. We'll be covering all things email marketing: B2B, B2C, newsletters, deliverability, email design, AI & more.
You can expect to walk away with new email strategies, the very latest digital trends, and how to step up your email performance. But don't worry, we also like to have fun. This year's theme is rom-com, so there will be DJs, meet-cutes, and a cutest pet contest. (Start prepping your dog's headshot now.)
Don't miss out. Join us Nov 12th & 13th for the largest virtual & free email marketing conference, powered by Constant Contact.
🧱 Around The AI Block
🤦♀️ Anthropic lost billions of dollars last year and said its AI could destroy humanity.
🤑 OpenAI's annualized recurring revenue nears $70 billion.
🏢 Meta now has a version of Muse just for small businesses.
🤔 Trump says working with Xi on AI would hurt U.S. companies.
🦾 Trump admin launches AI Gov site powered by Gemini, Grok.
🕵️♂️ Trump announces vague ‘morally binding’ AI deal among tech CEOs for ‘tremendous self-policing’.
💪 OpenAI’s latest features take direct aim at the app store model.
🤩 OpenAI expands ChatGPT’s plug-ins with app-like interfaces and automations.
🤣 The internet is convinced Elon Musk’s xAI trolled OpenAI’s ‘Dots’ launch.
👍 OpenAI gives Codex reusable cloud environments that work across devices.
👨⚖️ Anthropic says rogue AI agents pose uncertain legal risk for the company.
🛠️ Trending Tools
For the Smart Schedulers: Reclaim.ai is an AI calendar that doesn't just block off meetings; it automatically finds the best time for your habits (like gym or reading) and breaks, reshuffling them instantly if a high-priority task drops in. It’s the end of "scheduling crisis."
For the Curious Readers: ExplainPaper turns scary, brain-melting PDFs into “ohhh, that makes sense” explanations. You highlight the paragraph that broke you, and the AI rewrites it in plain English just like a tutor sitting right next to you. Seriously, it’s perfect for mastering new topics in half the time.
For the Vibe Seekers: Endel uses AI to generate real-time, personalized soundscapes that adapt to your heart rate, the weather, and even the time of day. Whether you need "Deep Work" focus or "Wind Down" sleep tracks, it builds the audio specifically for your body’s current state.
🤖 AI Workout Of The Day: How to Create Stunning 3D Animations from Text and Videos
While the big studios were busy arguing over render farms, AI just handed the keys to the kingdom to... well, everyone.
We’re talking "Pixar-level" visuals from a text prompt. If you’ve got an imagination and ten fingers (okay, one finger for clicking), you’re officially an animator.
The MVP: Krikey AI: If you want to go from "zero to hero" without a manual the size of a dictionary, Krikey AI is the runaway winner. It’s specifically designed to be so simple a 10-year-old could run a mini-studio from their tablet.
How to Build Your Masterpiece:
Hit Krikey.ai and choose a 3D avatar from the library or build your own from scratch
Type a prompt like "doing a backflip" into Quick Animate or upload a video of yourself to have the AI copy your moves.
Type your script into the Magic Studio to generate a voiceover that’s perfectly lip-synced to your character’s mouth.
Personalize the Scene: Pick a 3D background, tweak the lighting, adjust cinematic camera angles, and add text effects.
Hit export. Download it as an MP4 for the 'Gram or a GIF for the group chat or socials and you’re done.
Pro Tip:
Krikey AI adds AI Lip-Sync dialogue in 20+ languages.
For more cinematic, "moody" clips, creators are also flocking to Luma Dream Machine.
The Cost: Most tools offer free credits daily, so you can play before you pay!
💡 Prompt To Try:
Prompt: “A sleek superhero in a metallic suit flying down from the sky, landing on one knee in a powerful 'hero pose,' then slowly standing up and dusting off their shoulder with a smirk.”
Prompt: “A small, fluffy monster doing a high-speed breakdance routine, ending with a clumsy trip and a goofy wave to the camera while sitting on the floor.”
Prompt: “A stylish 3D character performing a smooth 'moonwalk' across the stage, followed by a celebratory backflip and a peace sign pointed at the lens.”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.
