When Brad Smith from Microsoft had called for the regulation of Facial Recognition technology in July I was concerned about where that might go, as it could easily result in stifling innovation. I was therefore relieved to see the principles that Microsoft put forth yesterday, which are for the most part quite sensible.
In particular, I agree and strongly support the due process suggestion on government’s use of facial recognition technology. Surveillance of an individual using facial recognition should require a court order.
I am also a big fan of requiring an API to enable third party testing. This is in fact the first instance I am aware of in which a large tech company proposes such a requirement which I have written about frequently before and is a central part of what I call “Informational Freedom” in my book World After Capital. A great approach here would be for an organization such as NIST to publish a reference data set against which all facial recognition systems could be tested.
The only section of the Microsoft proposal that I think is somewhat under specified and potentially subject to bad regulation is titled “Ensuring meaningful human review.” The goal of this section is laudable, which is to require a human in the loop for high stakes decisions instead of operating fully automated. But the criteria for when that might apply are broad and vague and could wind up encompassing a lot of the positive use cases. I would suggest limiting this part of the proposal to the exercise of government power.
Overall this is an incredibly thoughtful contribution from a technology leader to the discussion of how we can use our new capabilities for good.
