TLDR: New research from UC Davis computer scientists reveals that popular generative AI browser assistants, while offering convenience, pose significant privacy risks by extensively collecting and sharing sensitive user data with both first-party servers and third-party trackers for profiling and personalization.
A groundbreaking study by computer scientists at the University of California, Davis, has shed light on the alarming privacy implications of generative AI (GenAI) browser assistants. These tools, designed to enhance web browsing through features like summarization, translation, and question-answering, are found to be collecting and sharing vast amounts of personal and sensitive user data, raising serious concerns about digital privacy.
The research, led by Ph.D. student Yash Vekaria from Professor Zubair Shafiq’s lab, investigated nine widely used GenAI browser assistants: Monica, Sider, ChatGPT for Google, Merlin, MaxAI, Perplexity, HARPA.AI, TinaMind, and Copilot. The team conducted rigorous experiments to assess implicit and explicit data collection, as well as profiling and personalization practices. Their findings indicate that these assistants frequently gather sensitive information and transmit it to their own servers and external third-party trackers, such as Google Analytics, for the purpose of user profiling and personalized content delivery.
“These assistants have been created as normal browser extensions, and there is no strict vetting process for putting these up on extension stores,” Vekaria stated. He emphasized the critical need for users to be aware of the inherent risks, advocating for greater transparency initiatives to empower users in making informed decisions about their privacy. The study highlights that some of these extensions may even violate their own privacy commitments and potentially run afoul of U.S. regulations like HIPAA and FERPA by collecting and funneling health and student data.
Also Read:
- AI Agents’ Increasing Control Over Computers and Phones Raises Significant Security Concerns, Study Warns
- New Research Uncovers Critical Hijacking Vulnerabilities in Leading AI Agents
The paper, titled “Big Help or Big Brother? Auditing Tracking, Profiling and Personalization in Generative AI Assistants,” is slated for presentation and publication in the proceedings of the prestigious 2025 USENIX Security Symposium on August 13. The researchers underscore that addressing these pervasive privacy risks requires a collaborative effort across the entire GenAI ecosystem, rather than relying on a single entity to implement safeguards. The findings serve as a stark reminder that while AI-powered tools offer convenience, they often come with a significant trade-off in terms of personal data privacy, necessitating robust protective measures and increased user vigilance.
