AI Browser Assistants Found Collecting Sensitive User Data from Private Websites

Researchers have uncovered alarming privacy violations by AI-powered browser assistants, with nine out of ten tested tools capturing and transmitting highly sensitive user data. The study examined popular extensions including ChatGPT, Microsoft Copilot, and Merlin AI, revealing their ability to extract medical records, banking details, and even social security numbers from password-protected portals.

University College London's Anna Maria Mandalari warns these tools operate with unprecedented access to online activity, potentially crossing legal boundaries in their data collection practices. Only Perplexity AI emerged unscathed from the tests, which simulated real-world scenarios like online banking and medical record checks.

The findings raise serious concerns about the trade-off between convenience and privacy in the AI era. As browser extensions become increasingly sophisticated, their ability to intercept and decode sensitive information between users, corporate servers, and third parties creates new vulnerabilities in digital security.