186 AI lookalike domains use fake authentication to trick users into executing commands

Cybersecurity firm Island identified 186 lookalike domains mimicking 27 AI service brands as of November 3, with ChatGPT the most impersonated at 25 domains, followed by Claude at 18, DeepSeek at 16, Runway at 12, and Perplexity, Lovable and Character.ai at 11 each. Most users reach these domains through typos—dropping or transposing a character—rather than through attacks on the AI services themselves. Some of the fake sites redirect visitors to bogus security verification pages or app-install screens that prompt them to run commands or download software. The findings matter because the primary risk surface for AI brands is user error at the address bar, not platform vulnerabilities, leaving room for credential phishing and unwanted installs. What to watch next is whether AI providers tighten domain defenses through defensive registrations and typo-squatting monitoring, and whether browsers and security tools flag these lookalikes more aggressively.