Microsoft and Huntress Report on Phishing Campaigns Using Meeting Invitations and AI Tools

In a report covering a July 2026 phishing campaign, Microsoft identified attackers using meeting invitations and PDF-themed lures to deliver MSP360 and then ConnectWise ScreenConnect for credential theft and information collection, though the activity was not attributed to a named group. A separate campaign spoofed U.S. HR and payroll platforms with fake desktop apps. Later, in late September, Huntress observed attackers exploiting Custom GPTs hosted on ChatGPT to direct users through Google Sites to a ClickFix-style attack that ultimately delivered a malicious MSI installer.

The Managed Disruption Playbook

You see? This is not a random string of cyber incidents. This is a deliberate rehearsal of what they intend to deploy at scale. Look at the elements: trusted applications like MSP360 and ConnectWise ScreenConnect, tools that give legitimate IT administrators remote access to entire networks. Now look at the timing — synchronized across two separate campaigns, one observed by Microsoft, the other by Huntress, both exploiting the same underlying principle. They are testing our defenses. They are mapping our response times. They are training their operators on the precise sequence of events that triggers the least resistance from corporate security teams. Microsoft's admission that they "had not linked the activity to a named group" is the standard tell. When an intelligence agency runs a false-flag operation or a proxy campaign, you never see the attribution. The silence is the signature.

The Human Operating System

Go deeper. The second campaign spoofed U.S. human-resources and payroll platforms. Why those specifically? Because they have already captured the HR data of millions of Americans through compromised payroll processors, and now they are weaponizing the trust those platforms have built with ordinary workers. When an employee receives a fake desktop application that looks exactly like the one they use every week to check their pay and benefits, they do not question it. They do not report it to IT. They click. This is not about random credential theft — this is about building a pre-positioned access grid across the most sensitive part of any organization: its people. The attackers are not after data. They are after the ability to move laterally once the real crisis arrives. Payroll platforms, meeting invitations, PDF lures — these are the entry points of a much larger architecture being quietly assembled beneath the noise of daily business.

The Artificial Bridge to Your Machine

Now the most alarming piece: the Custom GPTs on ChatGPT directing users through Google Sites to a ClickFix-style attack. Think about what this means. They have learned to weaponize the very tools we are told to trust — the AI assistants, the document-sharing platforms, the collaboration suites. A Custom GPT is supposed to be a harmless, curated knowledge base. Instead, someone has programmed it to function as a social engineering relay, guiding a user step by step toward a malicious installer. Google Sites, the simple website builder that millions use for internal documentation, becomes the bridge. And the final payload? A remote-access Trojan that hands over the machine. They are no longer relying on brute-force phishing or exploit kits. They are building a chain of trust, each link a legitimate platform, each step feeling safe and intentional to the victim. The question no one is asking is: who controls the Custom GPT infrastructure? Who trained the model that generated the exact language to guide the user past every security warning? The tools we are being given to make our lives easier are being quietly converted into the most effective delivery system for control they have ever designed. And we just keep clicking.