Introduction
I've spent the last few years watching AI agents go from a fun demo to something that writes my code and drafts my emails. This week, something changed. On October 21, 2026, Microsoft Threat Intelligence reported that a China-aligned hacking group it tracks as Storm-0940 used an open-source AI agent framework called ARTEX to run a supply-chain attack against ConnectWise, then used the stolen ConnectWise credentials to attempt sign-ins to more than 2,250 Microsoft 365 tenants.
Not a human attacker at a keyboard. AI agents doing reconnaissance and privilege escalation at machine speed — in some cases, Microsoft says, going from target to privileged access in under three hours.
I build WordPress sites for a living, and my first thought was the same one you're probably having: this sounds like an enterprise problem. It isn't. The attack went through a trusted vendor — the kind of third-party app small businesses connect to Microsoft 365 and then forget about. If you run a site or rely on any SaaS tool that touches your business accounts, this story is about you.
What Actually Happened
ConnectWise makes software for managed service providers (MSPs) — the IT firms that look after email, backups, and cloud services for thousands of small businesses. Its cloud products (Cloud Backup, SaaS Security, and Cloud Management, formerly SkyKick) work by registering an application inside your Microsoft 365 tenant: you grant that app permissions once, and from then on it can back up mailboxes or manage users without anyone logging in. That convenience is exactly what the attackers exploited.
According to Microsoft, Storm-0940 — a group active since at least 2021 that focuses on data theft — got hold of the credentials for several ConnectWise application identities. How they were originally stolen has still not been made public. On October 15, 2026, the security firm Huntress spotted the stolen credentials being used to sign in to customer tenants, and in some cases the ConnectWise apps granted themselves administrator roles. ConnectWise disabled seven app identities and suspended Cloud Backup on October 17, taking SaaS Security and Cloud Management offline as a precaution.
Microsoft's contribution is the "how." The group drove the campaign with ARTEX, a freely available framework in which AI agents run an attack from reconnaissance to data theft with little human input. Using the stolen credentials, the agents attempted sign-ins to more than 2,250 Microsoft 365 tenants. Most attempts failed — the agents were trying their luck at scale. But in some cases it took under three hours from picking a target to holding privileged access.
A brute-force campaign ran in parallel: between October 10 and 14, an automated password spray from 144.172.64.99 hit 327 accounts at eight organizations with more than 151,000 failed sign-ins, peaking at 2,900 per minute — with no successful sign-in observed.
ConnectWise has since started a phased restoration, with backup and restore returning first through a new application. The company says existing backup archives remain encrypted, isolated, and preserved, with no indication they were compromised.
What Is ARTEX, and Why Is It Different?
ARTEX is the part of this story I keep coming back to, because it changes the math of cyberattacks for everyone. It is an open-source, multi-agent AI framework built for autonomous penetration testing — the kind of security testing where you hire someone to try to break into your systems before a criminal does. It's written in Go with a Next.js front end, and it can work with any AI model. It was published on GitHub on July 26, 2026, by a developer known as "Autumn-27." After CrowdStrike reported its use against South Korean banks in early October, the developer denied involvement and said the project would go closed source. The original repository has since been taken down — but forks and rebrands such as ScopeWeaver already circulate.
Let me be honest: ARTEX is not malware in the traditional sense. It's a tool with a legitimate purpose. The problem is what happens when a tool that automates the boring, time-consuming parts of an attack falls into the wrong hands. A human attacker used to need hours of manual reconnaissance and patient privilege escalation. ARTEX automates that pipeline: it finds a leaked credential, tests what it unlocks, and moves through cloud tenants — at a speed and scale no human team can match.
And this wasn't a one-off. Microsoft says several actors in different regions now use ARTEX: CrowdStrike attributed its VAPOR PANDA campaign against more than 200 organizations in 14 countries to it, the North Korean groups Onyx Sleet and Pearl Sleet have used it, and the financially motivated group Storm-3358 is using it against banks in Southeast Asia. As of November 4, Censys counted 1,465 publicly exposed ARTEX console hosts on the internet.
The Five-Minute Key Vault Heist
If ConnectWise shows what ARTEX does at scale, research published on November 5, 2026, shows what it does with a single leaked secret — and this is the part that worries me most as a developer. Microsoft Threat Intelligence and GitGuardian found ARTEX agents scanning GitHub for exposed credentials and automatically turning them into cloud access. In one case, an agent went from a GitHub search to pulling data out of an Azure Key Vault in five minutes.
The old advice was always "rotate it when you find it." That assumed a human would eventually stumble on your leaked key, and that you had some time. That assumption is now dead. ARTEX is purpose-built to find a leaked credential and immediately test what it unlocks. The agents found OAuth application secrets, signed in as service principals, and moved through Azure and Microsoft 365 tenants — automatically.
This is directly relevant to anyone who builds websites. How many of us have committed a config file with a real API key or pushed a .env file by accident? I have — and if you're honest, you probably have too. The old risk was that someone might find it someday. The new risk is that an agent finds it within minutes and tests it against every cloud service it can reach. Treat every secret as if it's already public the moment it leaves your machine, because functionally, that's now true.
Why This Matters for WordPress Site Owners
You might be thinking: I don't use ConnectWise or Microsoft 365 — why should I care? Three reasons.
First, the attack pattern is the threat, not the specific vendor. The ConnectWise compromise worked because a trusted third-party application held powerful permissions inside customer tenants, and nobody was watching what it did. WordPress sites work the same way: your backup plugin, your security plugin, your email marketing tool — every one of those is a vendor identity that could be abused the same way. The question isn't whether you use ConnectWise — it's how many third-party apps have standing access to your stuff, and when you last reviewed them.
Second, the speed changes the game for small sites. Attackers used to target big fish because manual attacks were expensive; now an agent can test thousands of tenants in hours, so your five-page brochure site or your client's WooCommerce store is exactly the kind of target that gets swept up — not because you're important, but because you're reachable.
Third, your clients' Microsoft 365 tenants are your problem now. If you manage websites for clients, there's a good chance those clients use Microsoft 365 for email, and their MSP may have used ConnectWise products. The checks below are a service you can offer your clients this week, too.
What to Check This Week
Here's the practical checklist I'm working through myself, drawn from Microsoft's guidance and the security community's write-ups. You don't need a security team — just an afternoon and admin access.
1. Review every third-party app connected to your Microsoft 365 tenant. In the Azure portal, check Enterprise Applications for apps holding admin-level permissions. If you or your MSP ever used ConnectWise Cloud Backup, SaaS Security, or Cloud Management (formerly SkyKick), make sure the old app registrations are disabled and only the new ConnectWise identity is active. Remove anything you don't recognise — and check for extra credentials on your service principals, since Microsoft warns an added credential can outlast the clean-up of the original app.
2. Check your sign-in and audit logs from October 1 onward. Look for directory listings from unusual IPs, mail or file downloads, forwarding rules, and delegated mailbox permissions you didn't create. Watch for the IPs 144.172.64.128, 144.172.64.99, and 144.172.64.60, and user agents like Python-urllib, python-requests, curl, and azure-toolkit — none of which should be signing into your tenant.
3. Audit your GitHub repositories and secrets today, not someday. Search your public repos for API keys, OAuth secrets, and connection strings, and rotate anything that was ever exposed — even if you "only" left it up for a few minutes. Five minutes from GitHub search to Key Vault is the new benchmark.
4. Lock down your WordPress admin and plugin access the same way. Remove stale administrator accounts, and rotate the API keys of any plugin you can't account for. If a plugin offers OAuth instead of a stored password, use it — OAuth tokens can be revoked centrally. And make sure your backups are actually restorable, because the ConnectWise story shows that even your backup vendor can become the attack path.
5. Turn on the alerts you keep meaning to turn on. The agents signed in through apps tenants had already authorized, so no user had to log in — user-focused alerts alone won't catch it. You need alerts on service principal sign-ins and application permission changes too.
Conclusion
The ARTEX story isn't really about one hacking group or one vendor. It's about a shift in who — or what — runs cyberattacks: a China-aligned group used an open-source AI agent framework to turn stolen ConnectWise credentials into sign-in attempts on more than 2,250 Microsoft 365 tenants, and the same framework went from a GitHub search to an Azure Key Vault in five minutes.
For two years, the conversation about AI in security has been mostly optimistic — AI will help defenders find vulnerabilities faster and spot attacks earlier. All of that is true. But the ARTEX story is the other side of the same coin: the same agent technology is now running attacks at machine speed, and the basics matter more, not less. Microsoft says the activity did not result from a compromise of its own systems. The weak point was a trusted vendor application.
For WordPress and Elementor site owners, the takeaway is simple and urgent: audit your connected apps, check your logs back to October 1, rotate any secret that ever touched a public repository, and lock down your admin access. The tools that protect you are the same ones you've always had — they're just more important now that the other side has agents doing the work.
FAQ
What is ARTEX?
ARTEX is an open-source, multi-agent AI framework for autonomous penetration testing, written in Go with a Next.js front end. Its agents can run an attack from reconnaissance through data exfiltration with little human input, and it works with any AI model. Published on GitHub on July 26, 2026, the original repository was taken down in October 2026 after threat actors abused it — but forks and rebrands continue to circulate.
What happened in the ConnectWise attack?
On October 21, 2026, Microsoft attributed a supply-chain attack to Storm-0940, a China-aligned group. Using ARTEX and stolen ConnectWise application credentials, the group attempted sign-ins to more than 2,250 Microsoft 365 tenants. Huntress spotted the activity on October 15, and ConnectWise disabled seven app identities and suspended Cloud Backup on October 17. In some cases the agents moved from targeting to privileged access in under three hours.
Was my Microsoft 365 tenant affected?
Only Microsoft and ConnectWise can confirm that for a specific tenant, but if your organization or your MSP used ConnectWise Cloud Backup, SaaS Security, or Cloud Management (formerly SkyKick), review your sign-in and audit logs from October 1 onward, check for the IPs 144.172.64.128, 144.172.64.99, and 144.172.64.60, and make sure the legacy SkyKick app registrations are disabled.
How did an AI agent steal data from Azure Key Vault in five minutes?
In research published November 5, 2026, by Microsoft and GitGuardian, ARTEX agents scanned GitHub for exposed credentials, found OAuth application secrets, signed in as service principals, and moved through Azure tenants. In one case the agent went from a GitHub search to exfiltrating data from an Azure Key Vault in five minutes. Treat any secret that reaches a public repository as compromised immediately.
What should WordPress site owners do about this?
Review every third-party app and plugin that holds credentials or admin access to your sites, remove stale admin users, rotate API keys you can't account for, prefer OAuth over stored passwords where possible, and make sure your backups are restorable. The same supply-chain logic applies: your security is only as strong as the weakest connected app.
