AI can infer personal attributes based solely on the patterns of advertisements individuals are exposed to, without requiring access to personal data or browsing history, according to researchers from UNSW Sydney and QUT. The study, which analyzed over 435,000 Facebook ads from 891 Australians, demonstrated that AI could predict traits such as gender, age, education, employment, political preference, and economic standing through ad exposure alone.
This method is reported to be over 200 times cheaper and 50 times faster than traditional human analysis of ad patterns. Researchers noted that even short browsing sessions can provide ample data for AI to construct an accurate personal profile.
Browser extensions were identified as potential threats, as they require permissions to read web content and could collect ad exposure data covertly. Popular extensions, like ad blockers and coupon-find tools, could misuse these permissions to silently gather ad data and send it to potential attackers.
This scenario poses a significant risk, as the stealthy operation of these extensions allows them to function normally while simultaneously invading personal privacy. The researchers emphasized that no hacking is required, and the advertising platforms remain unaware that their systems are being exploited for surveillance.
A VPN does not mitigate this issue, as ads can still reach the device regardless of the user’s internet connection. The study calls for privacy laws that not only address the data collected but also the inferences drawn from passive ad consumption. Researchers argue that individuals cannot completely opt out of the ad economy, making the need for legal protections around ad exposure critical.
“Your ad stream is a fingerprint that AI can now read,” the researchers stated, underscoring the necessity of evolving regulations to safeguard such personal information.
