Microsoft Threat Intelligence identified a malicious Chromium extension spoofing the Perplexity AI brand to deceive users into installation. The extension's main function was to intercept search traffic and collect data, including real-time keystrokes, using Manifest Version 3 capabilities and declarativeNetRequest rules. It routed full search queries and keystrokes through attacker-controlled infrastructure hosted on a typosquatted domain (perplexity-ai.online) before redirecting users to legitimate search providers. The extension also overrode browser default search settings and requested permissions inconsistent with legitimate AI assistants. Google removed the extension following responsible disclosure. This campaign illustrates how threat actors use AI-related branding as a social engineering vector.

The malicious extension compromises user privacy by capturing keystrokes and search queries, potentially exposing sensitive information. It hijacks browser search settings, redirecting traffic through attacker-controlled infrastructure, enabling data interception and possible further exploitation. The use of typosquatted domains increases the risk of user deception. Although no known exploits in the wild are reported, the data interception and keystroke capture pose a medium risk to affected users.

Google has removed the malicious extension from the Chrome Web Store following responsible disclosure. Users should uninstall any suspicious extensions, especially those spoofing AI brands like Perplexity AI. Organizations should strengthen user awareness training to recognize social engineering tactics involving AI branding. Implement layered security controls to detect and block similar malicious extensions and monitor for unauthorized changes to browser settings. No official patch is applicable as this is a malicious extension rather than a software vulnerability.