Synopsis
Privacy experts are flagging significant risks with Meta's AI glasses pilot for visually impaired citizens in Gujarat. Concerns centre on potential privacy violations for users and bystanders due to always-on cameras, and challenges under India's data protection law. While the technology promises independence, calls are mounting for stringent safeguards against data misuse and unauthorised access before deployment.Listen to this article in summarized format
They said the technology presents challenges under India’s Digital Personal Data Protection (DPDP) Act.
According to Meta, the glasses allow users to access AI hands-free to understand the world around them.
Advocate Apar Gupta, cofounder and director of Internet Freedom Foundation, said the glasses could capture homes, routines, people, documents and other sensitive information during daily use.
“Data captured so a blind user can read a medicine label should never flow into AI model training or advertising,” Gupta said.
He also flagged third-party integrations such as Be My Eyes, which Meta has integrated into the Ray-Ban Meta glasses.
Be My Eyes CEO Michael Buckley said more than 1 million blind or low-vision users globally use the platform, supported by more than 10 million volunteers across over 180 countries. He said live video calls initiated through the glasses are separate from photos and videos captured through AI features.
“All data is processed and stored in the US. That is clearly disclosed to users and volunteers when they join the platform, and they consent to it,” Buckley said.
He added that Be My Eyes does not review or annotate user videos but retains data indefinitely for safety, law enforcement requests and technical investigations.
Akshayy Nanda, partner, Saraf & Partners law firm, said AI glasses create a new privacy challenge because people nearby may be recorded without their knowledge. He said the government should identify places such as hospitals, schools and voting areas where recording should be restricted.
“The moment data captured through the glasses is shared with a third party, whether that is a cloud provider, processing pipeline or accessibility service, that third party is dealing with personal data in its own right,” Nanda said.
US-based lawyer and civil liberties activist Mishi Choudhary said the DPDP Act was not designed specifically for ambient, first-person AI capture.
“It has no wearable-specific rules for bystander notice, environmental capture, live computer vision, sensitive inferences and public sector pilots,” she said.
Choudhary said the pilot should prohibit facial recognition, advertising use, law enforcement access and AI training using beneficiary or bystander data.
“We can use the technology for our benefit without trading our rights,” she said.
Disability rights group The Accessibility Coalition (TAC) said AI glasses could significantly improve independence for people with disabilities. TAC secretary general Nipun Malhotra said accessibility and privacy should not be treated as competing goals, but safeguards around data storage, third-party access and government access must evolve alongside the technology.
However, Meta pushed back against the privacy concerns.
“Photos and videos are private to users. Humans review AI content to improve product performance, for which we get clear user consent,” a Meta spokesperson told ET in an email without responding to specific questions.
On May 25, Meta announced through a letter of intent that it would collaborate with the Gujarat government on AI initiatives, including accessibility through AI glasses. Details of the pilot are still being developed.
A Gujarat government official said the proposal would be examined by the relevant departments, including the science & technology department and the social justice & empowerment department. The departments did not respond to ET’s queries.
The privacy concerns come as Meta faces regulatory scrutiny in the UK and Kenya and litigation in the US over the Ray-Ban Meta glasses. A class-action lawsuit in the US alleges misleading privacy claims and inappropriate data-sharing practices, including concerns over human review of sensitive footage.
Experts have suggested several safeguards before deployment, including a public data protection impact assessment, transparent data terms, device-level processing wherever possible, strict limits on data retention, separate consent for AI training, accessible grievance mechanisms and independent audits.
Gupta of Internet Freedom Foundation said the pilot should not proceed without public consultation with beneficiaries. Saraf & Partners’ Nanda said authorities must clearly define accountability for data breaches and complaints, including the roles of data fiduciaries and processors.
