Scams & Phishing Hackers use fake OpenAI, SpaceX, and Anthropic event invites in new phishing scam to steal Microsoft and Google accounts

[Brownie2019]

Content source

https://www.techradar.com/pro/security/hackers-use-fake-openai-spacex-and-anthropic-event-invites-in-new-phishing-scam-to-steal-microsoft-and-google-accounts

• AI-themed phishing campaigns found using fake OpenAI, Anthropic, and SpaceX events to steal corporate credentials
• Attackers combine device-code phishing, reverse proxies, and MFA bypass techniques targeting Microsoft and Google accounts
• Researchers warn employee interest in AI tools is increasingly being exploited as a high-success phishing lure

If you suddenly get invited to an exclusive SpaceX event and even offered free access to Grok Heavy, kindly turn it down, as this is most likely part of a new phishing attack that’s been circulating around the web.

A report from security researchers Push Security found threat actors are using different AI-themed lures to get people to share their corporate login credentials.

As per Push’s report, threat actors were recently seen creating fraudulent websites for nonexistent events: an OpenAI Academy livestream, a private SpaceX briefing on AI marketing, an Anthropic presentation at AWS re:Invent, and similar.
People who landed on these pages were told they can get free goodies, as well: a free ChatGPT Pro subscription, access to Grok Heavy, and even shares of SpaceX. However, to claim these rewards, the victims were first told to log in - using either their Microsoft, or Google account.

And this is where the scam lies. Instead of actually logging in and getting the rewards, all victims would be doing is sharing their login credentials with the hackers.
"Employees want access to the latest AI tools, making them excellent bait for attacks," Push Security said in its analysis of the campaign.
Kali365 and VoidProxy
Push says threat actors are using two phishing-as-a-service (PhaaS) platforms in this campaign, called Kali365 and VoidProxy. Depending which route they choose (Microsoft or Google), they get served a different path, both ending in credential exposure.
Those that choose to continue with their Microsoft credentials trigger an attack through Kali365. The victim receives a device code and instructions to enter it on Microsoft’s legitimate device login page. Once the victim signs in and approves the request, Microsoft issues an authentication token to the attacker’s Kali365 session. The fake event page then redirects the victim to a fake Microsoft sign-in page operated by VoidProxy, which captures additional credentials and session cookies.

Those that choose “Continue with Google” trigger a different technique. Here, the victim is shown a reverse-proxied version of Google’s legitimate sign-in page, operated by VoidProxy. The proxy relays the authentication process while capturing the victim's password, MFA response, and session cookie, which allows the attackers to grab an authenticated session without needing to move past MFA.

Push doesn’t know exactly how the links are distributed. It’s safe to assume they’re being moved through phishing emails, social channels, or Telegram groups. What we do know is that the attackers are using Google and Zapier to shorten the links and redirect the visitors, in order to avoid being detected by security scanners and URL reputation systems.
Abusing the trends
"All of these detection evasion techniques, plus the MFA-bypassing AiTM and device code phishing payloads, mean this attack is designed to get around or through the controls that most organizations rely on to stop phishing attacks today," Push said.

It’s now been five years since the first version of ChatGPT was released, and since then the tool has captivated people from all walks of life, and was successfully used in countless phishing campaigns. In the past we’ve reported on campaigns involving Claude installation pages promoted through search advertisements, as well as malicious links disguised as legitimate AI content (shared conversations, downloadable artifacts, and custom GPTs).

In July 2026, ANY.RUN reported fake event invitations impersonating Meta, OpenAI, and Anthropic, hosted on legitimate Zoom event pages. More recently, Cofense observed a campaign revolving around fake Anthropic invitations distributed through GoTo Webinar.
As these tools get better, and as they penetrate deeper into every aspect of our private and business lives, they will be abused by threat actors more and more.

 

[Bot]

Hi Brownie, the particularly important lesson here is that seeing Microsoft’s genuine login domain does not necessarily make the process legitimate.

With the Microsoft route, the victim enters the supplied device code on Microsoft’s real site and completes normal authentication. The password need not be disclosed during that step; instead, the victim authorizes Microsoft to issue tokens to the attacker-controlled Kali365 session. According to Push Security’s analysis, this campaign then adds a VoidProxy page for another opportunity to capture the password and session cookie.

The Google route is adversary-in-the-middle phishing: VoidProxy relays the genuine sign-in process while stealing the password, MFA response and resulting session cookie. “MFA bypass” is convenient shorthand, but MFA itself has not been cryptographically defeated. The victim successfully performs MFA inside an attacker-mediated session, after which the attacker reuses the captured authenticated session.

A couple of evidence boundaries are worth retaining:

• Push observed and blocked visits to these fake event pages in customer browsers, so this is an observed campaign rather than a hypothetical technique.
• Push did not establish how the links were distributed. Email, social media and Telegram are plausible routes, not confirmed ones.
• The report demonstrates credential/token-capture capability, but it does not publish a victim count or success rate. Calling AI invitations attractive bait is fair; calling this particular campaign demonstrably “high-success” goes beyond the available figures.

For users, the practical rule is simple: never enter a device code supplied by an event, recruiter, support caller or unrelated website. Device codes are intended for signing into a device or application that you knowingly initiated. Also, go independently to the organizer’s official site rather than trusting an invitation link. A free SpaceX share should probably make the alarm bell ring without assistance.

If someone interacted with one of these pages:

• Merely viewing the invitation or event page is not evidence that the account was compromised by this technique.
• Entering the Microsoft device code and approving the sign-in should be treated as an account incident, even if no password was typed into a fake page.
• Entering Microsoft or Google credentials and completing MFA is also an account incident. Contact the organization’s IT/security team immediately, change the password from a trusted route, revoke sessions/tokens, and inspect sign-in activity, account recovery settings, connected applications, forwarding rules and mailbox rules. A password change alone may not invalidate every stolen session.
• Google Workspace administrators should reset the user’s sign-in cookies as well as the password; Google specifically recommends both for a suspected compromise.

On the administrative side, Microsoft recommends blocking device-code authentication wherever it is not genuinely required, ideally testing the Conditional Access policy in report-only mode first. Its current guidance is quite direct about getting as close as possible to a unilateral block. Phishing-resistant passkeys or security keys help substantially against reverse-proxy phishing, while device-code abuse needs that additional flow-specific control.

Sources

• Fake event invites combine Kali365 and VoidProxy payloads
• Block authentication flows with Conditional Access policy