Hackers use fake AI and SpaceX event invites to steal corporate logins

The lure of exclusive AI access

Security researchers at Push Security have identified a new phishing campaign that targets the high level of interest in artificial intelligence. Hackers are creating fraudulent websites for events that do not actually exist, such as an OpenAI Academy livestream or a private SpaceX briefing on AI marketing.

To entice victims, these fake pages offer high-value rewards for attending the nonexistent events. These perks include free ChatGPT Pro subscriptions, access to Grok Heavy, and even the promise of SpaceX shares. These incentives are designed to appeal to employees who want to stay ahead of the curve with the latest technology.

The primary goal of these lures is to get users to click a link and provide their corporate credentials. Once a user lands on the fake event page, they are prompted to log in using either their Microsoft or Google account to claim their rewards. Instead of receiving any goodies, the victims unknowingly share their login information directly with the attackers.

How the technical bypass works

The attackers use different technical methods depending on which login provider the victim chooses. For those using Microsoft credentials, the campaign utilizes a platform called Kali365. The victim is prompted to enter a device code on a legitimate Microsoft page, which then issues an authentication token to the attacker's session.

After the initial token is captured, the fake event page redirects the user to a different site operated by VoidProxy. This site is a fake Microsoft sign-in page designed to capture additional credentials and session cookies. By combining these methods, the attackers can gain full access to the user's account while bypassing standard security checks.

For users who choose to continue with Google, the attackers use a different technique involving a reverse proxy. This setup shows the victim a version of the legitimate Google sign-in page while the proxy relays the authentication process in the background. This allows the hackers to capture the password, the multi-factor authentication (MFA) response, and the session cookie all at once.

Evasion and distribution tactics

To ensure these phishing links reach their targets without being blocked, the threat actors are using clever redirection methods. They use services from Google and Zapier to shorten the links and redirect visitors. This helps them avoid detection by URL reputation systems and automated security scanners that organizations typically rely on.

While the exact distribution method is not confirmed, researchers assume the links are being spread through various channels. These could include phishing emails, social media channels, or private Telegram groups. By moving through these diverse platforms, the attackers can reach a wider audience of potential victims who may be less suspicious of the source.

The complexity of these attacks highlights a growing trend in cybercrime. Push Security noted that these payloads are specifically designed to get around the controls that most organizations use to stop phishing. By combining device-code phishing with MFA-bypassing techniques, the attackers can penetrate deeper into private and business environments.

A growing trend of AI impersonation

This campaign is part of a broader trend of exploiting AI-related hype for malicious purposes. In the past, researchers have seen campaigns involving fake Claude installation pages promoted through search ads. They have also seen malicious links disguised as legitimate AI content, such as shared conversations or custom GPTs.

Other reports have identified similar tactics where attackers impersonate major tech companies like Meta and Anthropic. For example, some campaigns have used legitimate Zoom event pages or GoTo Webinar links to host fake invitations. These methods show that attackers are constantly finding new ways to leverage trusted platforms to deliver their payloads.

As AI tools continue to penetrate every aspect of our lives, they become increasingly valuable targets for abuse. The fact that it has been five years since the first version of ChatGPT was released shows how quickly the landscape has changed. Threat actors are now able to create highly convincing lures that exploit the very tools people are trying to use for productivity.