In partnership with
Welcome Automaters, 👋
Ever scrolled through your feed, stumbled across a five-paragraph motivational speech about "key learnings," and thought to yourself: "There is absolutely zero chance a living human wrote this"?
Well, gather round, because LinkedIn just gave us the ultimate power move: you can now officially snitch on it!
On Thursday, July 30, the Microsoft-owned professional network rolled out a brand-new reporting option literally named "Seems like AI slop."
LinkedIn's Chief Product Officer, Hari Srinivasan, got surprisingly candid about the epidemic in a public post, admitting that generic chatbot text has become a top operational headache. His whole point? People open LinkedIn to connect with actual professionals sharing genuine experiences, not a bot making up fake life lessons on a Tuesday afternoon!
How LinkedIn Plans to Clean Up Your Feed
LinkedIn isn't just giving us a snitch button; they’re completely overhauling how they handle machine-written garbage behind the scenes.
Here are the key changes rolling out:
Training the Security Algorithms: Every time you click the new report button, the new button feeds data back into LinkedIn's detection classifiers so the system gets smarter at catching fake posts before they ever hit your feed.
Cracking Down on Bot Spam: The platform is already intercepting and blocking hundreds of thousands of automated comment attempts every single day.
The Gentle Warning Nudge: Instead of instantly banning people, LinkedIn will start privately notifying creators whose drafts sound overly robotic, giving them a chance to fix their tone before their peers flag them.
Killing Their Own Robot Ghostwriter: In a hilarious plot twist, LinkedIn is officially shutting down its controversial "Enhance your post" AI writing assistant! They’re swapping it out for a simple spellchecker, and proofreader effectively telling users: "Fewer AI ghostwriters, more of you just being you."
Now this entire push isn't happening in a vacuum. The internet is facing a full-blown spam crisis, and startups are dropping like flies because of it:
The Digg Tragedy: The revived Digg platform was forced to shut down its Reddit competitor back in March, openly admitting it simply could not keep up with the overwhelming wave of bots flooding its forums.
The Cloudflare Milestone: Web infrastructure giant Cloudflare recently confirmed that automated bot traffic has officially surpassed total human traffic across the entire web, hitting that dystopian milestone way faster than experts originally predicted!
The Detection Boom: Platforms everywhere are scrambling to build digital mops. Substack recently rolled out its own newsletter AI detector, while security startup Pangram just raised $9 million specifically to fight spam content.
The Bottom Line:
The dead internet theory is feeling a little too real lately, but platforms are finally fighting back. If you are still relying on ChatGPT to write your professional commentary, consider this your official warning: your connections are about to hit that report button!
But here's our two cents: AI was trained on human writing, which means it's gotten scary good at sounding human. So how exactly is LinkedIn planning to tell a "real perspective" from a really convincing AI one? And more importantly, how do they separate a false report from a legit one? Because let’s be real, current AI detectors are notoriously unreliable.
And right now, this button feels less like a solution and more like an invitation for chaos. Picture it: you write a genuinely great post, someone doesn't like your opinion (or your success), and suddenly you're getting flagged as "AI slop" simply because your grammar was too clean or your point landed too well.
Not trying to rain on LinkedIn's parade here. The intent is solid, nobody wants a feed full of robotic humble-brags. But good intentions and good execution are two very different things. So the question is, will this button build trust, or just give people a new way to be petty?
We'll be watching closely to see whether this actually cleans up the slop, or just adds a fresh layer of drama to everyone's notifications.
Here's what we have for you today
🤦♀️ Anthropic’s AI Models Hacked Real Companies Completely by Accident
So AI just had an absolutely unhinged week in cybersecurity. And we mean that in the most "it is giving both superhero and supervillain energy" way possible!
Let’s start with the drama.
Following OpenAI's recent security scare on July 21, Anthropic decided to conduct a proactive internal audit of its own cybersecurity evaluations.
And guess what?? The investigation uncovered a jaw-dropping revelation: three separate Claude models broke into the internet, and then into the live production systems of three different real-world organizations.
And the wild part? Nobody commanded them to do it!
The access traced directly back to a misconfigured testing setup run alongside a partner firm called Irregular. The environment was supposed to be completely isolated from the web, but a misunderstanding over whether the test setup had live internet access left a door wide open.
Anthropic made a point not to pass the buck, stating they’re taking complete ownership of the cleanup while Irregular conducts its own review. But because of that open connection, three different Claude models went rogue in very different ways:
Claude Opus 4.7 (The Persistent Hacker): The oldest model recognized it was messing with live corporate infrastructure in every single test run. But instead of freezing, it rationalized that the real company must be part of the test, then went ahead to swipe live access credentials, and poked around a production database anyway!
Mythos 5 (The Delusional Rogue): Spotted subtle signs that it was connected to the actual web, but successfully convinced itself it was still inside a simulation. It then published a malicious software package straight to the public Python registry (PyPI), which external systems actually downloaded and executed before anyone caught on!
The Internal Research Model: Anthropic's newest internal test model was the sole hero of the group, completely halting its own operations the moment it concluded the target was an authentic, live system.
Now, a single model stopping itself out of three isn't exactly a gold medal for machine ethics, but here we are!
Naturally, after the news broke, everyone began comparing this to OpenAI’s recent saga, but Anthropic pointed out key differences between the two incidents:
The Breakout Method: Where OpenAI's model actively exploited a software vulnerability to escape its sandbox, Anthropic's models simply walked through a door that had been mistakenly left unbolted.
OpenAI's Spree: OpenAI confirmed its model used exposed credentials across four accounts on four services, utilizing one as a staging ground and another for storage.
The Snitch Factor: Anthropic caught its own breach during internal reviews. In fact, two of the affected organizations had no idea their systems were even touched until Anthropic called to notify them!
Important Note: Anthropic noted that these test models ran without standard public safety filters, since these evaluations are built to measure raw, uninhibited model horsepower. Unsurprisingly, they are now calling for strict, military-grade isolation controls on all future AI security testing.
Now for the good news! While Anthropic was playing damage control, Google and Microsoft proved that AI can be the ultimate defense mechanism when pointed in the right direction. Here’s what the reports say:
Powered by models like Gemini, Google patched an astonishing 1,072 security vulnerabilities across just two Chrome versions in June alone. To put that in perspective, that single-month total beat the combined patches from the previous 23 versions over two whole years (which totaled 1,036)!
Riding the exact same momentum, Microsoft credited its internal AI tools with hunting down and fixing a record 570 flaws this month.
Meanwhile, Apple is sitting suspiciously quiet, logging a modest 482 patched bugs this year with no exponential AI curve in sight.
To summarize this wild week, here’s where we stand on AI in cybersecurity:
Defense Is Moving at Light Speed. Google and Microsoft prove that AI can catch bugs at a scale human engineers could never match, turning slow, manual patching into an automated, industrial operation.
Rare Corporate Transparency. Anthropic uncovered this mistake internally, owned it completely, and publicly reported itself, just as OpenAI did, setting a great standard for industry accountability.
Models Lack Reality Awareness. When an AI model assumes the real world is just "part of the game," the security risks stop being theoretical and start affecting real databases.
Raw Models Are Unpredictable. Running frontier models without public safety classifiers shows just how aggressive and persistent these systems get the moment guardrails come down.
The Bottom Line:
AI is officially the ultimate double-edged sword in cybersecurity. The exact same trait that makes it a world-class bug hunter, its ability to relentlessly pursue a goal without fatigue or second-guessing, is precisely what makes it dangerous when it misinterprets reality wrongly. The line between a digital locksmith and an unintentional burglar has never been thinner!
The best marketing ideas come from marketers who live it.
That’s what this newsletter delivers.
The Marketing Millennials is a look inside what’s working right now for other marketers. No theory. No fluff. Just real insights and ideas you can actually use—from marketers who’ve been there, done that, and are sharing the playbook.
Every newsletter is written by Daniel Murray, a marketer obsessed with what goes into great marketing. Expect fresh takes, hot topics, and the kind of stuff you’ll want to steal for your next campaign.
Because marketing shouldn’t feel like guesswork. And you shouldn’t have to dig for the good stuff.
🧱 Around The AI Block
👷♂️ Forward-deployed engineers are the AI industry’s latest talent obsession.
👉 Goldman Sachs asset arm forms an AI investing platform, memo shows.
⛔ China warns of retaliation if US sticks with robot ban.
🤖 Reddit reports a solid quarter but shows signs of AI’s impact.
🧑⚖️ Europe is preparing to hit Roblox and ChatGPT with its strictest platform rules.
👨⚖️ Judge says Trump admin still lacks evidence for Anthropic ‘supply-chain risk’ label.
🗣️ OpenAI's Sam Altman discusses rogue agent with US senators, as Trump considers AI controls.
👩🎓 AI Tutorials
Using ChatGPT to Generate 1000s of Leads in Any Niche (Free + Easy Method).
And: Top 27 ChatGPT Make.com Automations [Automate With AI].
So tell us, what’s the single most annoying, tedious task in your daily workflow that you desperately wish an AI could just handle for you?
Hit reply and the next video might just be around your exact problem!
Think You Know What AI Does Next?
Which model leads the next benchmark? Which AI lab ships the next major breakthrough?
Kalshi lets you trade on real-world AI and technology events as the industry moves. If you follow launches, model updates, and benchmarks closely, put that knowledge to work.
Bonus credit varies from $15 to $500. Terms apply.
🛠️ Trending Tools
For the Kitchen Wizards: Honeydew Recipe Manager combines meal planning, inventory tracking, and grocery list generation with features like household sharing and Instacart integration. It’s designed to help families stay organized, save money, and cut down on waste.
For the Smart Readers: Blinkist AI (Blinkist Connect) uses AI to generate "Key Insight" summaries of almost any book or long-form article you find on the web. You can forward a long PDF or a link to the app, and it will generate a 15-minute "audio blink" so you can learn the core concepts while you’re walking the dog.
For the Inbox Minimalists: SaneBox uses AI to analyze your past behavior and automatically filters "noisy" newsletters and distractions into separate folders, leaving only the high-priority humans in your main view. It’s the closest thing to "Inbox Zero" without actually trying.
🤖 AI Workout Of The Day: How to Control What AI Tells Your Teen: A Step-by-Step Guide to ChatGPT’s New Parental Controls
If you’ve ever worried that ChatGPT might be teaching your kid how to "jailbreak" their bedtime or (God forbid) giving them romantic/life threatening advice, we’ve got some spicy updates for you. OpenAI has a massive toolkit to help you keep the guardrails up.
Step 1: The Invitation
You can't just spy in secret. You have to link accounts.
The Move: Open ChatGPT on your phone or browser. Click your Profile Icon: Settings: Parental Controls.
The Link: Hit "Add family member" and enter your teen’s email or phone number.
The Catch: Your teen has to actually accept the invite on their device. Once they do, they’ll pop up under your "Family Members" list.
Step 2: Set the "Quiet Hours"
Teens love a late-night rabbit hole. But you can stop those 3 AM "How to build a cold fusion reactor" sessions right here.
In your Parental Controls menu, select your teen's name and toggle "Quiet Hours" to ON. Set your start and end times (e.g., 10 PM to 7 AM). During this window, ChatGPT will essentially go to sleep for them.
Step 3: Lock Down the "Spicy" Stuff
OpenAI defaults to higher safety for teens, but you can double-down on the specifics:
Reduce Sensitive Content: This aggressively filters out graphic, violent, or "mature" responses.
Disable Voice Mode: If you don’t want them having long, spoken heart-to-hearts with a robot, you can turn this off.
Kill Image Generation: Worried about "edgy" AI art? Toggle this OFF to disable DALL-E 3 on their account.
Step 4: The "Brain Wipe" (Privacy First) 🧠
One of the coolest (and creepiest) features is Memory. If you don't want the AI building a "personality profile" of your kid, toggle it to OFF. This ensures the AI "forgets" everything once the chat ends.
Also: make sure to opt-out of "Model Training" so their private rants aren't used to train the next version of GPT.
Pro Tips for the Digital Parent:
Transparency is Key: Your teen will be notified when you change a setting. There is no "stealth mode" parenting here.
The "Unlink" Alert: If your teen gets naughty and unlinks their account to hide something, you’ll get an immediate email or text alert.
Emergency Safety: If the AI detects your teen is in a mental health crisis, it’s designed to trigger an "Acute Distress Alert" to your phone.
The Privacy Balance: You still cannot read their actual chat logs. OpenAI keeps those private to build trust with the teen, but you can see the general settings that are active.
💡 Prompts to try: General Marketing Strategies
You are an experienced marketing strategist. I will provide details about a business, and you will create a tailored marketing strategy. Please include:
. Key messaging and brand positioning – How should the brand be perceived and communicated?
. Best marketing channels (digital + offline) – Which platforms and mediums will have the most impact?
. Campaign ideas for customer engagement – Creative ways to attract, convert, and retain customers.
. Suggested pricing and promotional tactics – Recommendations for competitive yet profitable positioning.
. Steps for short-term wins and long-term growth – Quick actions plus sustainable strategies.
Format the response in clear sections with bullet points and examples.
Whenever possible, give reasoning behind each recommendation so I can understand not just the ‘what’ but the ‘why.’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.
