🚨 BREAKING: The FBI just issued a FLASH alert about ARTEX, an open-source AI pentest framework that hackers are using to automate attacks.
Some intrusions went from first scan to privileged access in under 3 hours.
What it is:
❋ multi-agent tool written in Go, you give it a goal in plain English
❋ it doesn't ship its own model, it plugs into Claude, GPT, DeepSeek, Kimi, GLM
❋ attackers mostly run DeepSeek Flash models: cheap, and they show no guardrails when the request is framed as a pentest
Where it's been used:
❋ China-aligned Storm-0940 used it in the ConnectWise Cloud Backup compromise, then tried to sign in to 2,250+ Microsoft 365 tenants
❋ personal data of ~67,000 people stolen from South Korean financial institutions
The dev pulled the GitHub repo in October. Forks, mirrors and a paid "ARTEX PRO" are still out there, and researchers found 1,600+ exposed ARTEX consoles.
What to check this week:
❋ list every internet-facing app, especially partner portals, mobile backends and forgotten side apps
❋ scan repos, env vars and JS bundles for hardcoded secrets
❋ put admin panels (Kudu/SCM, backups) behind private access
❋ rate-limit portals and alert on sequential record lookups
❋ if your servers make LLM API calls you didn't expect, investigate
73
OpenAI just launched SEO for ChatGPT and almost nobody has noticed yet.
It's called Business Hub.
Claim your company and you can see:
- How many times ChatGPT mentions you
- How many clicks it sends to your site
- The topics of conversations where your site gets cited
- Your info, prices and stock, pulled from a product feed you control
1.2 billion people use ChatGPT every week. Until now you had zero visibility into what it said about you.
If you run a business, claim your business today. Takes 10 minutes. Shopify stores can connect straight from the admin.
The brands that start reading this data now will know what to fix months before everyone else.
1
2
125
🚨 In October 2025, the best AI agent could finish 1 in 40 real freelance jobs.
Now it's 4 in 10.
That's Claude Fable 5.2 on the Remote Labor Index: 240 paid Upwork projects across 3D, CAD, architecture, video, audio and data work.
Human experts compare each deliverable with what the freelancer actually delivered.
The previous Claude model scored 17.9%. The next best model today is at 21%.
The test leaves out client calls and anything you only see results from months later, like SEO. Those are now the hardest parts of the job to automate.
2
96

