Workers training artificial intelligence behind Ray-Ban Meta Smart Glasses say they regularly see highly private footage captured by the devices — including people leaving bathrooms naked, changing clothes, and other intimate moments recorded unknowingly by users.
“In some videos you can see someone going to the toilet, or getting undressed. I don’t think they know, because if they knew they wouldn’t be recording,” one worker said.
The claims were revealed in an investigation by Swedish newspapers Svenska Dagbladet and Göteborgs-Posten, which interviewed data annotators in Nairobi working for outsourcing firm Sama, a contractor helping train AI systems for Meta Platforms.
The report highlights the hidden human labour behind AI technologies and raises fresh questions about privacy as wearable devices collect increasing amounts of personal data from users’ everyday lives.
Why Data Annotators Regularly See Highly Private Smart Glass Footage
Artificial intelligence systems depend on massive amounts of data to learn how to interpret images, voices and human behaviour. Despite the term “machine learning,” much of this training still relies on human workers reviewing and labelling data so algorithms can understand what they are seeing.
These workers — known as data annotators — often operate in outsourced centres in lower-income countries, where they classify images, review videos and check whether AI systems provide accurate responses.
However, as consumer AI devices such as smart glasses become more common, the data collected is increasingly drawn from people’s daily lives — including homes, workplaces and private conversations.
“I think few people would want to share the details of their daily lives to that extent. But when it is presented in a fun and appealing way, it becomes harder to see the risks”, says Petter Flink, an IT and security specialist at IMY.
Experts say many users do not realise that interactions with AI systems may be reviewed by humans, raising concerns about privacy and transparency.
How Meta Uses Outsourced Workers To Train Ray Ban AI Systems
Ray-Ban Meta Smart Glasses are designed to function as an AI-powered assistant that can capture photos and videos, answer questions, translate languages and interpret what the wearer sees in real time.
During a product presentation in Menlo Park, California, Mark Zuckerberg showcased the glasses as a potential successor to smartphones — a wearable device capable of integrating artificial intelligence into everyday activities.
The glasses connect to the Meta AI system, which processes user commands and visual data through cloud-based infrastructure.
To improve accuracy, AI systems are trained using interactions with users, including voice commands, images and sometimes videos. Some of this data may be reviewed automatically or manually to help the AI recognise objects, understand speech and generate better responses.
Workers at companies such as Sama help perform this training by labelling images, reviewing conversations with the AI assistant and verifying whether responses are correct.
Meta says such interactions are processed according to its terms of service and privacy policy, which state that some interactions may be subject to human review.
“When live AI is being used, we process that media according to the Meta AI Terms of Service and Privacy Policy,” Joyce Omope, a Meta spokesperson, said.
Balancing African Digital Job Growth With Ethical AI Privacy Standards
The investigation illustrates how the rapid expansion of the AI industry is creating new digital jobs across the world, including in Africa.
Countries such as Kenya have become important hubs for data annotation, employing thousands of workers who help train the AI systems used by major technology companies.
If the sector continues to grow, it could provide significant employment opportunities for African workers, connecting them to the global digital economy.
However, the report also underscores the ethical challenges associated with AI development, including worker exposure to sensitive content and concerns over how personal data collected by smart devices is used.
Addressing these issues through stronger data protection frameworks, clearer user consent and better safeguards for workers could help ensure that the growth of AI technologies benefits both users and the global workforce supporting them.