Blending virtual reality with artificial intelligence has the potential to revolutionize our experiences, but it also raises concerns about privacy. Recent research has shown that by analyzing the movements of individuals wearing virtual reality headsets, machine learning models can accurately predict personal data such as height, weight, age, and even marital status. This highlights how AI can guess sensitive information without users explicitly disclosing it.
The University of California, Berkeley conducted studies that demonstrated the accuracy of this prediction. In one study, researchers were able to identify a specific person from a pool of over 50,000 VR users with an astonishing 94% accuracy. They achieved this result by analyzing just 200 seconds of motion data. In another study, researchers successfully determined a person’s height, weight, foot size, and country with over 80% accuracy using data from 1,000 participants playing the popular VR game Beat Saber. Furthermore, personal information like marital status, employment status, and ethnicity could be identified with more than 70% accuracy.
To analyze this data, researchers employed a machine learning model that examined the data uploaded to virtual reality headsets, such as eye and hand movements. According to a Bloomberg report, lead researcher Vivek Nair explained that certain characteristics like age, gender, ethnicity, and country were relatively easy for the model to predict. For example, the model could estimate someone’s age based on their reaction time in hitting a virtual target. Faster reaction times are often associated with better eyesight and younger age. However, the model could also make educated guesses about income level, disability status, health status, and even political preference. This demonstrates the potential breadth of personal information that AI can infer from VR data.
The implications of these findings are significant as virtual reality headsets capture data that is not typically available through traditional websites or apps. This includes a user’s gaze, body language, body proportions, and facial expressions. Jay Stanley, a senior policy analyst at the American Civil Liberties Union, highlights that this convergence of VR and AI intensifies existing privacy concerns and introduces new ones.
Meta Platforms Inc., a major player in the VR industry, has been utilizing machine learning to augment its understanding of users. However, the extent to which VR data is incorporated into their algorithms remains unclear. In 2021, Apple’s privacy policy changes restricted the amount of data Meta could track on iPhones, resulting in a significant loss of revenue. This forced Meta to invest in AI and subsequently experience double-digit revenue growth this year by improving its ability to predict user preferences.
Meta has been running limited ads in VR headsets since 2021 but assured users that it would not use data processed and stored on the devices for ad-targeting purposes. When questioned about their data policy derived from headsets, Meta directed inquiries to its Quest Safety Center. Here, users can control their avatar, profile picture, name, and username privacy settings. Meta also clarifies that data sent to and stored on their servers will be dissociated from user accounts once it is no longer needed for service provision or improvement of eye-tracking features.
Meta’s previous collection of sensitive personal data has drawn regulatory scrutiny. In 2021, the company shut down its facial recognition system and deleted over 1 billion facial images. Biometric data like facial images are particularly sensitive as they are unchangeable and can uniquely identify individuals. Similarly, VR headsets capture equally sensitive data. However, because VR technology is relatively new, users and regulators may not fully comprehend its implications, making it potentially more dangerous.
Building privacy controls for VR headsets poses challenges compared to websites or apps due to the necessity of collecting data such as eye and hand movements for optimal functionality. Encryption of collected information or limiting stored data are potential solutions according to Jay Stanley. However, headset manufacturers face incentives to gather user information for marketing purposes. Researchers note that user awareness of the extent of data collection by VR headsets remains low. Combined with the power of AI predictions, it may be unrealistic to expect consumers to effectively protect their privacy in this context.
As the integration of virtual reality and artificial intelligence progresses, it is crucial to address the privacy concerns raised by these technologies. Striking a balance between providing immersive experiences and safeguarding personal data is paramount. Policymakers, industry leaders, and users must work together to establish clear guidelines and robust privacy controls to ensure that the benefits of VR and AI can be enjoyed without compromising personal privacy.