The Meta Ray-Ban glasses and Indian law
1. India is not likely to find success by regulating hardware because (i) such regulations will be hard to enforce in practice and in principle, (ii) the regulatory regime could set up a new license raj, and (iii) the regulations will age very poorly when similar technological capabilities come to earbuds, contact lenses, etc. along with new affordances.
2. India's data protection regime needs to address/tighten its existing rules in five areas:
(i) Meaningful notice — The DPDP Act focuses on the relationship between the data fiduciary and the data principal whereas AI wearables introduce a third-party problem in the form of people who are recorded incidentally or by accident. So the law should clarify that companies selling AI wearables have a duty to minimise data collection when that data pertains to bystanders, and introduce aggressive technical measures such as filtering faces out, automatically deleting non-user data, and where possible have on-device data processing (i.e. before it’s uploaded to a server).
(ii) Limited purpose — audio or video collected for one purpose — e.g. answering a user's question about a building — shouldn’t automatically become training data for future AI models unless users, and where feasible all affected individuals, have each consented individually.
(iii) Data retention — AI wearables work best when they have as much context as possible but at the same time governments should resist creating permanent ‘lifelogs’ of their wearers. Instead companies should justify why each wearable needs to retain particular categories of recordings beyond short periods (a few seconds).
(iv) Transparency obligations — All manufacturing and/or promoter companies should disclose — in plain language rather than obfuscatory legalese — what data the wearable processes locally and what it uploads, how long a data packet will be retained on-device and on the cloud, who can access it, and whether it will contribute to training or improving AI models. The state should independently audit each of these claims.
(v) Sensitive environments — Regulators should impose stricter obligations on whenever AI wearables are used in hospitals, schools, courtrooms, polling booths, research labs, and/or workplaces handling confidential information. Many of these places already regulate photography on their premises.
3. For two centuries or so now, cameras have been conspicuous and social norms have evolved around that being the case. So when someone raises a phone and points it at you, you infer that they may be recording you. AI-enabled glasses fit with cameras and voice detectors break that convention and, in a new turn, render the act of recording visually ambiguous. Put differently, there’s now an information asymmetry as the wearer always and with certainty knows whether the wearable is recording whereas everyone else is uncertain.
For this reason, governments have a reasonable basis to require an observable indicator of whether recording is occurring (provided it’s also technologically difficult to disable). A small external LED that lights up when a wearable is recording, and not just when it’s starting/stopping to record, might be the least intrusive solution. A startup chime or shutter sound alone is less effective because people can miss it, it doesn’t indicate whether recording continues, and could come in the way of legitimate uses such as recording a lecture with permission or documenting police interactions. Also, even if India requires indicators on all Ray-Ban Meta glasses henceforth, similar capabilities will soon appear in devices imported by private entities or which are made and sold by other manufacturers. This means the regulations need to be capability-agnostic, i.e. any wearable that can covertly record should provide a standardised and externally visible recording indicator.
4. Stronger indicators are more valuable when people are less likely to seek consent before recording — however, the law shouldn’t be measured against assumptions about the prevailing national culture. India has an arguably la-di-da approach to the concept of privacy and civility vis-à-vis informing someone before recording them. So a recording indicator can preserve a longstanding social cue that the new technology has removed and regardless of whether it is deployed in India or Europe. More broadly, technologies shouldn’t be allowed to displace established social expectations — but at the same time states shouldn’t stifle innovation. And the aforementioned instruments, including a more explicitly defined set of bystander rights to complement the newfound ambiguity of recording, are the way to go.
