A phishing campaign detected in late October 2026 uses fake event invitations for OpenAI, SpaceX, and Anthropic to steal work credentials and authentication sessions from Microsoft and Google accounts, according to research published by browser-security firm Push Security.
The campaign combines two distinct phishing-as-a-service (PhaaS) platforms—Kali365 for Microsoft device-code phishing and VoidProxy for adversary-in-the-middle (AiTM) attacks on Google accounts—to bypass multi-factor authentication. It targets employees with lures that promise access to AI tools or equity, exploiting strong workforce demand for cutting-edge software.
Push Security first detected and blocked visits to the fake registration pages on customer browsers in late October. The pages advertised an "OpenAI Academy" livestream, a private SpaceX briefing on "SpaceX AI for Marketing," and an Anthropic session at AWS re:Invent, Push reported1. Each asks the visitor to sign in with a work account to claim a reward: three months of ChatGPT Pro, three months of Grok Heavy, or a "SpaceX share."
The attack chain begins with an email invitation containing a short link from Google or Zapier, which redirects through a Cloudflare challenge designed to block automated security scanners. Signing in with Microsoft triggers a device-code attack: the page displays a code and instructs the victim to enter it on Microsoft’s real device-login page. Once approved, tokens are issued to the operator's Kali365 session, and the victim is then passed to a VoidProxy Microsoft sign-in page for a second credential capture. The Google path opens a VoidProxy reverse proxy that captures the password, MFA response, and session cookie in real time.
The operator built the event-themed lure pages using a kit Push tracks as MSDRoot, which first appeared on October 22. Push linked this kit to an earlier campaign kit called ZoomMock, which was used from May to October 2026, because both were hosted on the same primary domain and used similarly themed AI-branded events to target marketing staff. The research notes that the operator left artifacts from prior lures, including URLs referencing “Claude,” in the code for the OpenAI page.
The domains used to host the lures are rotated frequently. The eventrsvp[.]app domain was three days old at the time of investigation. Push stated that its behavioral detection identifies the Kali365 and VoidProxy payloads in the browser in real time, so even fresh domains are blocked without relying on blocklists.
The research does not include details on the number of targeted organizations or the success rate of the campaign. Push Security said its customers do not need to take any additional action.