626/69 Friday, November 13, 2026
Cybersecurity company Island has disclosed a campaign involving 186 fake domains impersonating websites and Artificial Intelligence (AI) services from 27 different brands. The campaign, named TypoPrompting, targets users who mistype website addresses or visit domains that closely resemble legitimate services. The attackers impersonated popular AI platforms including ChatGPT, Claude, Google Gemini, Microsoft Copilot, Grok, DeepSeek, Perplexity, Midjourney, and Cursor. Using typosquatting techniques, the threat actors redirected users to malicious websites and deployed three primary attack methods: ClickFix techniques that trick users into executing commands, malicious software capable of modifying browser settings, and browser extensions that redirect user searches to attacker-controlled search engines. Such activity could expose search queries and sensitive organizational information to unauthorized third parties.
Island researchers found that ClickFix attacks typically involved redirecting users through multiple websites while collecting browser information and public IP addresses before presenting a fake Cloudflare verification page. Victims were then instructed to copy and execute PowerShell commands on Windows or shell commands on macOS. In some cases, users were redirected to download archive files containing instructions to extract and execute malicious files. The infrastructure used to profile visitors included more than 211 initial domains, some of which impersonated banks, insurance companies, and enterprise software services. Researchers also identified eight fake domains impersonating AI services such as Claude, DeepSeek, Copilot, and ElevenLabs that redirected users to fraudulent security-check pages and prompted them to install a program called Net Guard. Although presented as a tool for checking Wi-Fi and DNS security, the software was capable of executing attacker-controlled JavaScript, changing the default search engine, modifying browser profiles, and installing additional Chrome extensions. Researchers also found that the MSIX file used in the campaign had the same hash as a legitimate Net Guard application available through the Microsoft Store.
Island also identified the fake domain preplexity[.]ai, which impersonated Perplexity and redirected users to a browser extension called Quick Search available through the Chrome Web Store. The extension could change the browser’s default search provider to find.flashysearches[.]com through its Extension Manifest without requiring executable JavaScript or additional permissions. As a result, users’ search queries could be sent to a third-party provider and potentially expose internal organizational information such as internal hostnames, customer names, or project details. Researchers identified 16 browser extensions associated with this activity, including GPT for Chrome, which showed approximately 80,000 users. Together, the extensions displayed a combined user count of approximately 972,000 as of November 3, 2026. However, this figure does not mean that all listed users had their data stolen. The campaign highlights the risks posed by fake domains impersonating AI services to distribute malware and manipulate browser search traffic. Users should verify URLs before accessing AI services, avoid executing commands presented by untrusted verification pages, and carefully assess browser extensions before installation. Organizations should also establish approved channels for accessing AI services to reduce exposure to these threats.
Source: https://thehackernews.com/2026/11/186-fake-ai-domains-redirect-users-to.html