Fake AI Event Invitations Target Microsoft and Google Accounts

By Tom Whitfield

Push Security detected and blocked visits to fake OpenAI, SpaceX and Anthropic event-registration pages in late October 2026, exposing a two-platform credential-theft chain built for corporate accounts. The pages required corporate sign-ins and combined two phishing-as-a-service platforms in one chain.

The pages closely resembled legitimate event websites and offered free access to premium AI services and other rewards, including a SpaceX share. A work-account sign-in became the price of registration.

Three offers defined the lure:

• An OpenAI event titled “ChatGPT for Social & Digital Marketing Teams” promised three months of ChatGPT Pro.
• A private SpaceX event on November 12 offered three months of Grok Heavy or a “SpaceX share.”
• An Anthropic event at AWS re:Invent 2026 used the same registration format, although Push Security found that page already offline.

Each page required visitors to sign in with a work account to register and claim the reward. Push Security found two options on the form: “Continue with Microsoft” and “Continue with Google.”

Inside the attack chain

Two custom phishing kits, ZoomMock and MSDRoot, generated the fake event pages and directed visitors to Kali365 and VoidProxy. The researchers believe the same threat actors run both kits because they share infrastructure and target the same employees.

How the links reached staff remains unclear. Current pages use links from trusted services, including Google and Zapier URL shorteners. Earlier versions spread through GoTo Webinar’s email systems or legitimate Zoom event pages. The report did not quantify victims, geographic reach or financial loss.

A Cloudflare CAPTCHA appeared before each registration page and read “Confirm you’re not a robot.” The gate prevents automated security platforms from scanning the site before the credential flow reaches the visitor.

The Microsoft and Google paths

Selecting “Continue with Microsoft” made the page contact Kali365, a phishing-as-a-service platform that generates a Microsoft device authentication code. The visitor copied that code and entered it on Microsoft’s legitimate device-login site, where they authenticated and approved the request. Kali365 then stole the resulting tokens for the victim’s Microsoft Entra account.

The attack abuses the legitimate OAuth 2.0 device authorization grant flow, so the victim need not disclose a password directly. After device-code authentication succeeded, the page redirected the user to a VoidProxy Microsoft login page, where attackers could also capture the password and session cookie. Push Security noted that attackers have used device-code phishing for data theft and cyberespionage before.

Those tokens can reach Microsoft 365 and connected single-sign-on applications. Push Security listed Salesforce, Google Workspace, Dropbox, Adobe, SAP, Slack, Zendesk and Atlassian among the exposed services.

For Google accounts, “Continue with Google” opened a popup containing a Google login page supplied by VoidProxy. The platform operates as an adversary-in-the-middle reverse proxy between the victim’s browser and Google, proxying login forms while collecting credentials, MFA responses and session cookies.

The two PhaaS services can therefore defeat multifactor authentication in different ways. Kali365’s device-code route bypasses MFA entirely; VoidProxy uses a captured authenticated session and can also intercept MFA responses. Push said both platforms let attackers bypass multifactor authentication.

A pattern, not a one-off

ANY.RUN documented a related campaign in July that abused legitimate Zoom event pages to promote fake Meta, OpenAI and Anthropic events. Those lures led to the same device-code and adversary-in-the-middle phishing methods.

Cofense also reported phishing emails impersonating Anthropic event invitations. GoTo Webinar’s email infrastructure distributed those messages, which redirected recipients through multiple websites to fake registration pages.

Push Security’s advice for employees is straightforward: do not enter Microsoft device authentication codes supplied by unfamiliar sites, and treat event invitations that demand corporate sign-in for rewards with caution. Organizations can restrict device-code authentication where they do not need it and use phishing-resistant MFA, including FIDO2 security keys, to help prevent session hijacking.

What does this change for security teams?

Identity teams now need to treat a branded event reward as an OAuth request, not harmless lead generation. A shortener, a Cloudflare CAPTCHA and a legitimate Microsoft login page can sit in one chain, making domain reputation and password awareness insufficient on their own.

The practical control is to limit device-code authentication, require phishing-resistant MFA such as FIDO2 keys, and monitor sign-ins that follow unexpected invitation links. The lure is the event; the payload is the corporate session.