AI Glasses Unveil Privacy Risks: An Interview with the Creators of I-XRAY

In an era where technology’s reach seems boundless, privacy concerns have taken center stage. Two Harvard students, AnhPhu Nguyen and Caine Ardayfio, have turned the spotlight on the vulnerabilities we face with their project. I-XRAY is an artificial intelligence tool that, at first glance, is a set of unassuming glasses. Within seconds of someone’s face coming into the view of the glasses, the tool can retrieve personal details like home addresses and phone numbers, and overlay these personal details onto the glasses screen using augmented reality. While the project is not being released to the public, its implications are profound, demonstrating how today’s AI, facial recognition, and data aggregation tools can invade privacy in ways that few of us imagined.

Nguyen and Ardayfio sat down with HTR to discuss their motivations, the ethics behind their project, and the growing necessity for digital privacy protections.

What began as a side project soon evolved into something much more significant. “I started building this for a birthday party of mine,” Nguyen recalls. “Like, we thought it’d be interesting as a demo.” As they showed it to more people, they realized that “it was a huge security and privacy issue.” 

The technology itself leverages a suite of advanced tools. Using discreet smart glasses like Meta’s Ray-Bans, reverse facial recognition engines like PimEyes, and large language models (LLMs), I-XRAY can automatically identify a person’s name, occupation, home address, and even partial social security numbers—all within less than a minute. Large language models enhance text processing abilities, as “once you have LLMs, you can not only get the name, but you can also search for that home address after you get the name, and LLMs also filter through … and logically figure out, oh, this person must be like around their 20s or 30s and 40s.” The scary part is that it’s all automatic. Once they tie these technologies together, it becomes disturbingly easy to retrieve someone’s personal information just from a photo.

Despite the technical success of I-XRAY, Nguyen and Ardayfio quickly recognized the risks of their creation. The more they tested it, the more they realized it wasn’t just a fun project—it had real consequences, Nguyen admits. Rather than market the tool, they decided to take the opposite approach and not release it. They want to put “guardrails in place so that it’s not…used unsafely.”

The purpose of I-XRAY is to demonstrate the current capabilities of smart glasses, AI, and public databases, and to raise awareness of just how exposed our personal data is. Ideally, “millions of people have… seen this video, and.., I would guess, that probably…1000s of people hopefully have…taken action, on clearing their data and will be more privacy conscious.” They provide a detailed guide to concerned viewers on how to remove their information from databases online here.

Nguyen and Ardayfio’s concerns are not unfounded. The pipeline behind I-XRAY taps into technologies already available to the public. PimEyes, for instance, is a reverse facial recognition tool that scours the web for matching images of a person’s face. Once a match is found, data aggregation services like FastPeopleSearch step in to supply names, addresses, and phone numbers from publicly accessible records.

While European regulations have restricted the use of real-time facial recognition, no such protections exist in the United States. For Nguyen and Ardayfio, the hope is that projects like I-XRAY will spark discussions about stronger data privacy regulations. Awareness has already increased, as they report receiving comments from a “lot of people [that] are quite surprised that they’re just on this company’s website where anyone can search this for free or for 30 bucks a month.”

Both Nguyen and Ardayfio are no strangers to ambitious projects. Since meeting as Harvard freshmen in MakerSpaces, the pair has worked on a slew of creative, tech-driven endeavors, including an electric skateboard controlled by hand gestures and a punch-activated flamethrower.

Nguyen describes his own learning style as a hyper “project based learning approach.” Together, they estimate that they’ve both undertaken around 30 technical projects.

Despite the technical achievements of I-XRAY, the project remains just one of many for the duo. “We didn’t even consider starting any company out of this,” Ardayfio admits. It’s “just an interesting build.” Instead, both creators are already moving on to new endeavors, with Ardayfio focusing on industrial tech and Nguyen pursuing a special concentration in human augmentation and exploring human-computer interfaces through AI and extended reality (XR). They’re considering building another startup together. 

As for the future of AI-powered surveillance tools, both Nguyen and Ardayfio acknowledge that commercial applications are inevitable. “ClearView is a real company that does this exact thing,” Nguyen points out. It’s only a matter of time before others catch on.

But the question remains: can the average citizen protect themselves from such invasive technology? Right now, opting out of these databases is a complex, frustrating process. Ardayfio suggests “making these big companies make an easy opt out” option so that people can remove their information from online databases.. Ideally, awareness of tools like I-XRAY will push for easier ways to opt out of these services. It shouldn’t be so hard to protect your privacy.

At the end of the day, tech like this isn’t going away. People need to take action themselves by removing their data where they can. And governments need to step in with regulations that prioritize our privacy.

For now, I-XRAY serves as a wake-up call—a glimpse into a future where our faces can unlock more than just our smartphones. It’s a reminder that, in an age of limitless technological innovation, privacy may be the next frontier worth protecting.

Author

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

You don't have permission to register

Discover more from Harvard Technology Review

Subscribe now to keep reading and get access to the full archive.

Continue reading