TL;DR
The ICO’s chief executive, Paul Arnold, has used a signed blog to set out how the regulator is thinking about AI wearables. He condemns harmful uses of smart glasses but argues the bigger problem is that privacy transparency in physical spaces relies on things people can see. A consultation on the ICO’s new code for AI and automated decisions is planned for 2027.
A regulator writing from experience
Arnold is severely visually impaired, and he opens with that. Glasses that combine a camera, AI and audio help him and others make sense of their surroundings with more independence. The ICO’s regulatory sandbox has also worked with Cross Sense, whose glasses for people with dementia won the Longitude Prize in March.
He is equally blunt about misuse, saying he is “shocked and appalled” by some of it, and that the stigma now attached to responsible users is a harm in its own right. That follows reports of people being filmed without knowing it, and a campaign to ban Meta’s AI glasses from the Tube.
The argument
Arnold’s central point is that CCTV signs, QR codes and a phone held up to film all work as visual cues. For people with little or no sight, those cues are often unavailable. Wearables make capture far less visible for everyone, so a problem that used to affect a minority now affects the whole public. His question is how anyone can object to data collection they cannot detect.
He is also frank about the limits of data protection law. Recording for purely personal use falls under the household exemption, so the law does not stop someone filming in the street. Posting that footage publicly probably falls outside the exemption. The rest, he says, is a matter for social norms, product standards and possibly new laws targeting specific misuse, not for the ICO alone.
What it means for UK organisations
For businesses deploying cameras, voice capture or wearables, our reading is that signage built for sighted passers-by is starting to look like an incomplete answer. Arnold wants privacy by design treated as a source of innovation rather than a compliance box, and he notes that some users already try to disable safeguards manufacturers build in. Firms selling or deploying these devices should expect questions about how the people being recorded find out.
Looking forward
The regulator becomes the Information Commission on 30 September, and this blog reads as an early marker of its priorities. The 2027 consultation on its code for AI and automated decisions is the concrete next step. Watch whether it deals directly with transparency that does not depend on sight.