Researchers uncovered a human-run phishing platform that pretends to be ad-management products for popular AI brands (ChatGPT, Gemini, Claude, and others). Victims are pushed to click a “Connect” flow that opens a convincing fake login window (showing trusted domains like accounts.google.com) to steal passwords and MFA codes in real time. The stolen access is then used to take over advertising/manager accounts, add attacker admins, and monetize the accounts by running ads or selling them.
A human-operated phishing platform built a set of convincing lookalike products impersonating AI advertising tools tied to brands like ChatGPT, Gemini, Claude, Perplexity, Meta Muse, and Manus. Each fake product had its own branding, pitch, and sign-in flow, but all shared a single call to action: a prominent Connect button. Clicking it triggered a browser-in-the-browser (BitB) window, a fake login popup drawn inside the page that displayed trusted-looking addresses such as accounts.google.com or an Okta tenant, even though the real browser never left the phishing domain. One identified site, museads.ai, pitched itself as an AI ads manager offering campaign optimization and spend audits to lure victims into connecting their business ad accounts.
The realism of the fake login window was central to the scheme's success. Because the address bar shown to the victim matched a familiar, trusted domain, many users would have little reason to suspect anything was wrong. Victims are assessed to have been driven to these landing pages through fake invitation emails impersonating trusted AI brands, adding a layer of legitimacy before the victim ever reached the fake portal. Behind the scenes, the platform recorded every password attempt, fingerprinted the device, and let a human operator choose which MFA challenge to present next, allowing real-time interception of one-time codes rather than relying on a static, easily-detected phishing kit.
The likely targets were advertising agencies, specifically media buyers, marketing staff, and manager-account administrators with access to ad platforms. The goal was not simple credential theft for its own sake but account takeover: gaining control of accounts with established, clean spend histories so they could be monetized directly through ad spend or resold.
Organizations should treat unsolicited connect-your-account invitations as high risk and verify sender and destination domains before clicking. Teams should be trained to recognize that legitimate sign-ins open in the browser's real address bar, not inside a page element. Since MFA codes can be captured and relayed in real time, phishing-resistant authentication methods are recommended. Finally, monitoring advertising account admin changes helps catch takeovers early, since attackers typically add their own administrators and downgrade legitimate owners after gaining access.
Mirage safely runs attacks like this one against your own team, so you find out what happens before a real adversary does.
It is a fake login window drawn inside a webpage that mimics a real sign-in popup, showing a trusted-looking address like accounts.google.com while the actual site stays on the attacker's phishing domain.
The attackers aim to take over ad manager accounts with clean spend histories, then add their own administrators and monetize the accounts by running ads or selling access.
Not fully. The platform recorded password attempts and let an operator choose which MFA challenge the victim saw, capturing codes in real time, so phishing-resistant authentication is recommended.
Victims are assessed to be driven to the landing pages through fake invitation emails impersonating trusted AI brands like ChatGPT, Gemini, and Claude.
You get an email: “Invitation to connect your AI Ads portal (beta)” for ChatGPT or Gemini. Looks legit, right? You click through to a site like museads.ai, promising AI campaign optimization. Front and center is a big “Connect” button. Click it, and a fake login window pops up inside the page showing accounts.google.com or your Okta tenant. This is a browser-in-the-browser, or BitB, attack. Every password try is logged, and the operator even chooses which MFA challenge you see, so they can steal codes in real time and add themselves as admins to your ad accounts. Aha test: if a Google, Meta, TikTok, or Okta login box appears inside another page, stop. Don’t log in there, open a fresh tab and go to the provider’s site directly.
Researchers report NovaCookies, a phishing-as-a-service toolkit that steals Microsoft 365 session cookies by proxying real logins in real time. The campaigns abuse genuine DocuSign email notifications to deliver a malicious document link that ultimately leads to an attacker-controlled Microsoft 365…
This weekly roundup highlights multiple real-world social engineering threats, including fake IT help-desk phone calls that push employees to phishing sites to steal passwords and one-time authentication codes. It also describes credential-phishing sites impersonating WhatsApp and Instagram that…
Researchers showed how a malicious web page could trick OpenAI’s Atlas AI-enabled browser into taking actions a user didn’t intend, like spamming WhatsApp contacts or modifying an Amazon account. The attacks used prompt-injection style instructions hidden in a seemingly legitimate “newsletter…
This weekly roundup includes two real social-engineering-driven incidents. Attackers used social engineering to access Apollo Global Management’s cloud platforms and steal sensitive personal data, and separately attempted a ShinyHunters phishing attack against ReliaQuest using a fake SSO login page…
Attackers are sending fake ChatGPT billing emails that pressure people to “update payment” within 48 hours to avoid service interruption. The message links to a convincing look‑alike ChatGPT login page via a Google redirect, aiming to steal OpenAI credentials.
The article describes real intrusions where attackers impersonate a company helpdesk and lure employees into "passkey, MFA, or SSO setup" steps. Victims are sent links via text (often to personal phones), leading to account takeover through adversary-in-the-middle phishing or device-code…
