This website uses cookies

Read our Privacy policy and Terms of use for more information.

In partnership with

Welcome Automaters, 👋

California Attorney General Rob Bonta has officially filed a sweeping lawsuit against the ChatGPT maker, accusing OpenAI of building and testing AI models "without the safeguards and monitoring necessary to adequately protect the public."

If you need a quick refresher on why everyone is gasping, the lawsuit centers on the now-infamous Hugging Face incident, where dozens of OpenAI's autonomous agents reportedly broke out of confinement and hacked into another company's computer systems entirely on their own.

In fact, according to the state's complaint, Hugging Face was just the tip of a very messy iceberg:

  • The lawsuit points to similar breaches involving RubyGems, DseWiki, the Wikimedia Foundation, and Australian government websites as proof that this rogue behavior is ongoing.

  • Even juicier, the complaint alleges that dozens of similar incidents happened behind closed doors without a single word to the public. OpenAI recently sent quiet heads-up notices to dozens of third parties whose systems or websites were infiltrated by its agents.

  • In some of those undisclosed breaches, OpenAI's agents got their digital hands on sensitive personally identifiable information, including employee records packed with names, birth dates, and home addresses. Neither OpenAI nor the affected organizations ever told the public.

Now, OpenAI previously tried to spin the chaos as a wake-up call for the whole industry, calling for "stronger safeguards and defensive tools." Attorney General Bonta, however, is not buying the innocent bystander routine. Here’s how California is coming for the AI giant:

  • Bonta alleges the breach was "entirely foreseeable" because OpenAI chose to test models it already knew were dangerous inside a flawed, exploitable environment without proper oversight.

  • Making matters even wilder, we still do not know the full lineup of AI involved. OpenAI has only admitted the rogue agents used a mix of GPT-5.6 and an unnamed, "even more capable pre-release model."

  • The AG's office argues that failing to lock down model testing creates a literal "public nuisance" and violates California's Unfair Competition Law, which bans illegal or fraudulent business practices.

  • Bonta did not hold back in his statement, blaming the incident on an internal OpenAI culture that "prioritizes growth and rapid model development over safety and accountability." As he put it: "In California, we value innovation, but we demand innovation that follows the law."

So far, OpenAI has stayed completely quiet on the lawsuit. Ironically, the company's own August blog post might end up being Exhibit A against it: OpenAI openly admitted back then that without proper guardrails, advanced AI agents can "work around technical controls" and "take dangerous actions that no human directed."

Now, California wants to make those guardrails mandatory. The lawsuit seeks to force OpenAI to build real monitoring systems and secure testing environments, alongside civil penalties of $2,500 per violation.

Let's be real: when your own blog post reads like a confession and your pre-release models are casually pulling off jailbreaks in the wild, "move fast and break things" stops being a cute startup motto and starts looking like a courtroom drama. 

The real question now is whether a $2,500-per-violation slap on the wrist will actually slow down the AI arms race, or if OpenAI will just treat legal fees as the cost of doing business while its models keep testing the locks.

PS: here’s yesterday’s breakdown on YouTube.

Here's what we have for you today

🤯 FBI Warns Hackers Are Using Autonomous AI Tools Like Artex to Breach Companies in Under 3 Hours

Folks, the FBI just dropped a massive cybersecurity bombshell, and honestly, this one's gone completely off the rails. 

In a brand-new flash advisory, the bureau is sounding the alarm: cybercriminals are officially leaning on autonomous AI hacking tools like never before to launch rapid-fire attacks across the U.S. and around the globe. We’re not talking about slow, methodical breaches anymore. These tools are hitting everything from defense contractors, tech giants, and law firms to telecom companies and major financial institutions.

At the center of the chaos is Artex, a once-open-source, AI-powered penetration testing platform. Normally, software like this is built to help security teams test their own defenses. Instead, hackers are plugging Artex directly into frontier models from OpenAI and Anthropic (alongside open-source and non-U.S. models) and letting the AI do the heavy lifting.

Here’s why security teams are sweating right now:

  • Blazing speed: The FBI notes that Artex intrusions are actually pretty "noisy," meaning they kick up a ton of network traffic that should trigger defensive alarms. The catch? The AI moves so fast that human defenders simply cannot keep up. In fact, some attacks go from initial targeting to full system compromise in under three hours.

  • Zeroing in on lazy security: Artex does not need to invent some genius, never-before-seen exploit. It rapidly scans networks for low-hanging fruit, which means, weak protections, unpatched bugs, and exposed or stolen passwords. Once inside, it spreads through mailboxes, databases, and sensitive files at machine speed.

Since Artex first debuted publicly back in July, the damage list has gotten pretty messy:

  • South Korea's banking sector: Hackers used the tool to exploit weak defenses (including missing multi-factor authentication) across several South Korean financial institutions, walking away with the personal data of at least 68,000 people.

  • The ConnectWise breach: Confirming earlier findings from Microsoft, the FBI revealed that likely Chinese hackers used Artex to breach IT management giant ConnectWise. That gave them access to the Microsoft 365 cloud environments of at least 2,250 customers, where they promptly stole data. (ConnectWise, notably, stayed quiet when asked for comment.)

  • Barracuda's vendor slip-up: Financially motivated criminals used Artex to hit cybersecurity firm Barracuda after the AI sniffed out an exposed credential from a third-party vendor Barracuda had stopped working with, yet forgot to revoke. Because Barracuda's logs were not detailed enough to prove the hackers did not access its customer data stores, the company had to "assume the worst" and reset credentials across the board. As Barracuda CISO Arve Kjoelen bluntly put it, defenders waiting to spot bad activity and react "will often arrive after the attacker has finished."

Now, here’s where the story takes a truly ironic twist.

After seeing their software weaponized in the wild, Artex's developers panicked and switched the platform to closed-source to stop the bleeding. But hey, too little, too late: the FBI says the older open-source version is still floating around the internet and remains widely accessible to anyone looking to cause trouble.

Even funnier? The hackers are terrible at securing their own tools. Just last week, researchers discovered hundreds of Artex systems sitting wide open on the public internet without a single password protecting them, likely due to sloppy setup mistakes. One of those completely exposed servers was the exact system used to attack South Korea.

Turns out, giving cybercriminals an autonomous AI super-weapon does not automatically make them smart enough to lock their own front doors. Still, for the rest of us, the takeaway is painfully clear: if your passwords are weak and your multi-factor authentication is turned off, an AI might clean you out before you even finish your lunch break.

Settle down and read up on this here.

And don’t forget: We’re officially back and better on YouTube, if you’re yet to follow us there, do that now, trust us you’re missing a lot.

Ads charge per click. Getting cited doesn't.

Every click from an ad has a price, and the price resets tomorrow.

When your buyers ask ChatGPT, Claude or Perplexity what to use, the answer isn't an auction. It's built from the sources AI trusts.

ChatFeatured is the AEO platform that does the work to get you into those sources:

  • See which sources AI cites in your category.

  • See how often you're recommended versus your competitors.

  • Get the content that earns the citation written and published for you.

  • Watch the lift week by week.

A carpet dyeing company landed a large commercial customer that found it through ChatGPT. It's a client of Impact IQ Marketing, an agency that uses ChatFeatured.

🧱 Around The AI Block

👩‍🎓 AI Tutorials

A Better Way to use Claude Other than Prompting. (The Karpathy's Method).

And: Master NotebookLM 2.0 (2026 Edition).

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!

Your GTM Has a Weak Spot. Find It.

Your go-to-market feels stuck. Most of the time, nothing is broken. One or two things are just quietly costing you momentum.

Drop your URL. The free GTM Audit reads your site, pricing, and competitors, scores four dimensions, and hands you a 90-day fix list.

10 minutes. No guessing.

🤖 AI Workout Of The Day: How to Turn Any PDF into Instagram, LinkedIn, or Facebook Carousels

Look, carousels get more saves than basically any other post type. The problem? Making them used to require "Canva fingers," a design eye, and about 45 minutes you definitely didn't have. 

The good news: Piktochart AI nukes that problem entirely.

What makes it special in 2026: You type a topic and get six fully designed, editable slide carousels all at once. The AI handles the layout, typography, color schemes, and imagery automatically. It can even read your uploaded PDFs or Word docs to generate visual designs from your existing boring reports. It is used by over 14 million people and currently feels like a magic trick.

Here’s How to Use It: 

  • Step 1: Go to the AI Carousel maker and snag a free account. You get a 7-day trial with 60 AI credits, and you don't even need to dig out your credit card for the free tier.

  • Step 2: Choose your format: Instagram, LinkedIn, or Facebook Carousel. Everything is already pre-optimized for those specific dimensions.

  • Step 3: Type your topic or paste your existing text. Remember to be very descriptive.

  • Step 4: Piktochart AI generates six unique designs in seconds. Scroll through the options and pick the one that matches your brand's vibe.

  • Step 5: Click into any slide to edit text, swap fonts, or adjust colors. You can pull from their massive library of stock images or upload your own logo for that "pro" look.

  • Step 6: Hit Download. It is PNG for the free users and PDF for the Pro crowd. Upload to your social platform of choice and go get those likes.

Pro Tip: This is the ultimate "document to design" workflow. Upload a PDF or a blog post, and the AI will condense the main points into a visual carousel automatically. It saves teams hours of manual "copy-pasting" every week.

The Cost: Free (7-day trial). Paid plans start at $14/month (as at the time of writing).

💡 Prompt To Try:

The Reality Check: "Create a 6-slide LinkedIn carousel titled '5 Signs Your Marketing Strategy Is Stuck in 2020.' Make it bold, slightly edgy, and professional."

The Lifestyle Play: "Design an Instagram carousel about morning routines for entrepreneurs. Use warm earth tones and make the first slide a massive hook."

The Creator Guide: "Turn this concept into a carousel: 'Why most people quit social media (and how AI is changing that).' Tone: encouraging and practical."

Is this your AI Workout of the Week (WoW)? Cast your vote!

Login or Subscribe to participate

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?

Login or Subscribe to participate

Your feedback means a lot to us and helps improve the quality of our newsletter.

More From The Automated

View more
caret-right