Island reports phishing platform posing as AI advertising tools

Island researchers described a human-operated phishing platform disguised as AI advertising products, which impersonated Gemini, Claude, ChatGPT and Perplexity.[1] One example was Muse Ads, presented on museads.ai by September 16, after Meta introduced Muse as a personal AI agent on September 8, 2026.[1] Muse Ads promised to help advertisers reach buyers, connect an advertising account and run sponsored placements.[1] The lures also included a Google Ads briefing, manager-account features and support for linked clients.[1] The date shown on Island’s report is October 6, 2026.[1]

Clicking Connect opened a browser-like window drawn inside the phishing webpage rather than opening Google.[1] That imitation could display trusted-looking addresses such as accounts.google.com while the actual browser remained on the phishing domain.[1] The fake browser was designed to adapt to Windows, macOS, iOS and Android.[1] The platform also fingerprinted devices using information including IP address, location, screen size and WebGL data.[1] For teams, the practical lesson is to check the real browser address and verify a purported AI service or connector through the vendor’s official website before granting account access.[1]

The platform supported authentication workflows for Google, Meta, TikTok and Okta.[1] It locally rebuilt provider interfaces and collected credentials and MFA state through its own APIs.[1] An operator could see each submission and decide what the victim saw next; operators could also reject an entry, prompt the victim to try again and preserve every submitted value.[1] That human control means an apparently familiar sign-in flow should not be treated as proof that a login is genuine.[1]

Island said the lures targeted agency staff, media buyers and manager-account administrators because advertising accounts are spending accounts.[1] Stolen accounts could be used to run attackers’ campaigns or sold for profit, and Island cited research that aged accounts with clean spending histories could sell for two to four times the price of new accounts.[1] Attackers could also add their own administrators and downgrade legitimate owners, with recovery taking weeks or months.[1] Island recommended phishing-resistant authentication, including origin-bound passkeys and hardware-backed authentication, to reduce reliance on reusable passwords and one-time codes.[1]