The Netherlands’ data protection authority, Autoriteit Persoonsgegevens (AP), has issued a blunt recommendation: do not buy AI-powered smart glasses. The watchdog’s investigation into six devices currently available on the Dutch market revealed that the integrated AI assistants are prone to dangerous hallucinations, misinterpreting critical data ranging from emergency medical instructions to legal boundaries.

Hallucinations in High-Stakes Scenarios

The AP’s testing uncovered failures that would be catastrophic in real-world applications. In one simulated emergency requiring CPR, a pair of glasses repeatedly assured the user that an ambulance had been dispatched, despite no contact being made with emergency services. Another device provided inappropriate dating advice to an adult expressing romantic interest in a minor, highlighting a severe lack of contextual understanding. These errors extended to reading medicine package instructions and interpreting severe weather warnings, proving that current AI models cannot yet be trusted with life-safety information.

Security Flaws and Privacy Concerns

Beyond algorithmic failures, the investigation exposed significant security vulnerabilities. The AP found that some devices allowed nearby attackers to connect without permission, view recordings, or remotely operate the camera. The watchdog also challenged manufacturer claims regarding recording indicators, noting that on some models, the lights signaling active recording could be easily obscured with tape without disabling the function. This ease of covert recording raises serious consent issues, particularly in healthcare settings or public spaces where filming is restricted.

The Meta Factor and EU Data Export

While the AP declined to name the specific brands tested, noting only that prices ranged from under €100 to over €500 and originated from different continents, the findings arrive amid growing scrutiny of major players like Meta. Meta recently released a camera-less version of its glasses, a move commentators interpreted as a response to privacy backlash. However, the AP’s investigation did not primarily focus on data protection regulations but rather on the reliability of the algorithms themselves. It did, however, warn that data recorded by these devices is often exported outside the EU, adding another layer of complexity to the compliance landscape.

Key Takeaways

  • The Dutch AP recommends against purchasing AI smart glasses due to unreliable AI advice in medical and legal contexts.
  • Testing revealed security flaws allowing unauthorized remote access and easy bypassing of recording indicators.
  • Data exported by these devices often leaves the EU, raising cross-border privacy concerns.
  • Manufacturers like Meta have responded to privacy backlash by releasing camera-less variants, though reliability remains the core issue.

The Bottom Line

For developers and consumers alike, this is a clear signal that AI integration in hardware is currently outpacing reliability. Until we see robust verification layers for AI outputs in safety-critical contexts, these gadgets are more liability than utility.