Lisbon, Aug. 18, 2025 (Lusa) - University researcher Anna Mandalari warns that the collection of personal data by Artificial Intelligence (AI) assistants in search engines jeopardises privacy and could be used for data breaches or online profiling.
Speaking to the Lusa news agency, the author of a study carried out by University College London in the UK and the University of Reggio Calabria in Italy said that "many search engine AI assistants have unprecedented access to users' online activity, including private areas such as health portals, banking and email".
"They often collect and share confidential information with their own servers or even with third parties, sometimes without the user's knowledge or consent," the academic added.
In this sense, the collection of diverse information such as names, email addresses, health records, location, income or browsing histories creates "significant privacy risks. Once collected, this data can be used for targeted advertising, profiling or data breaches."
The study in question reveals that several assistants transmitted full web page content, including information visible on screen, to their servers.
One of the AI assistants, Merlin, captured form entries such as bank or health data, while ChatGPT, Copilot, Monica and Sider could infer user attributes such as age, gender, income and interests, and used this information to personalise responses even across different browsing sessions.
Anna Mandalari says that to prevent this type of "theft", it is necessary to implement a "privacy by design" approach, where AI assistants process data locally on the device whenever possible, ask for the user's explicit and informed consent for any data collection and limit the retention of personal information".
"While regulations are essential, they are not enough on their own," she emphasised, adding that "real progress will come from cooperation between AI companies, internet service providers and regulators, combined with greater user awareness."
One of the ways to increase this awareness is, for example, real-time indicators showing what data is being collected and clear explanations of how it is used, leading users to make more informed choices.
"When privacy is recognised and promoted as an added value, rather than an afterthought, the market will reward companies that protect their users and change behaviour," she explained.
PYR/ADB // ADB.
Lusa
