The coalface of surveillance tech

Kashmir Hill's book blows the lid off a business model that smashes data-privacy, and delves into its evolution and deployment by states in recent years

Your Face Belongs to Us: The Secretive Startup That’s Dismantling Your Privacy

Author: Kashmir Hill

Publisher: Simon & Schuster

Pages: 352

Price: Rs 799

In November 2019, the author, who is a technology reporter at The New York Times, received a leaked corporate memo from a contact. This was written by Paul Clement, who had served as Solicitor-General of the US in the George W Bush administration.

Mr Clement was representing Clearview AI, a company she had not heard of. He claimed Clearview had developed face-recognition algorithms which could identify anybody whose picture was publicly available with 99 per cent accuracy, given one snapshot. He also claimed 200 law-enforcement agencies and police departments were using Clearview’s services.

She was sceptical. (She was also pregnant, which turned out to be important.)  Mr Clement refused to talk. Clearview AI had a near-blank website. The corporate headquarters was apparently just two blocks from the NYT. But the address didn’t exist. Linkedin searches turned up just one employee — she later learned he also didn’t exist.

When she spoke to contacts in different police departments, they enthusiastically endorsed Clearview. As a test she asked them to run her picture. One said there were no matches. This was weird — a vanilla Google Image search turns up dozens of images of Kashmir Hill. Another policeman said he immediately received a phone call from Clearview asking why he had “run this Kashmir Hill lady from the NYT?”

Clearview had two investors. One was tech billionaire and prominent Trump-backer, Peter Thiel, who ignored her attempts to get in touch. The other was a small New York-based venture capitalist. When she went to that office, the founder, David Scalzo, was sufficiently moved by her pregnant condition to speak to her. By January 2020, her persistence had convinced Clearview that she would write about it, come what may. So, the company instituted damage control and met her.

Two of Clearview’s founders had over a dozen meetings with her. One was coder Hoan Ton-That, a flamboyant individual with a Vietnamese father and an Australian mother. The other was Richard Schwartz, a NY wheeler-dealer who had been a key aide of Rudy Giuliani.  Clearview’s third founder, Charles Johnson, was a right-wing fake-news troll and holocaust-denier, whom Messrs Schwartz and Ton-That were trying to side-line. Mr Johnson was also willing to speak in detail to Ms Hill.

Ms Hill wrote an explosive article about this mysterious entity. That blew the lid off a business model that smashed data-privacy. It was the genesis of this book. Here, apart from dealing in detail with Clearview AI and its evolution, she provides a history of face-recognition research. 

Mr Ton-That was a computer science dropout. He had been a model, and played classical guitar and chess. Despite mixed ethnicity and fluid gender identity (he claims he’s “currently male but previously went by other gender identities”), he had right-wing, anti-immigration views. He had created invasive third-party apps on Facebook before getting into face-recognition along with his erstwhile friend, Johnson. Their networks connected to mainstream Republicans .

Mr Ton-That reviewed the public-domain research on AI, neural networks and face-recognition. He trawled the Internet scraping publicly available pictures. Clearview’s algorithm is independently rated as 99.76 per cent accurate in identifying people, including folks wearing sunglasses, Covid masks, and so on. (There have been misidentifications and that’s not surprising since we’re talking in very large numbers.)

Apart from enabling state-surveillance with all its ills, face-recognition can help criminals target people, and it leads to context and location data that can be used to identify residences, workplaces, family, friends and travel schedules. Ms Hill says Facebook, Google et al pulled back from face-recognition research when they realised the implications.

But Mr Ton-That stumbled almost by accident on an escape clause that lets him operate without going to jail, or being shut down with huge fines, (the company is fighting multiple lawsuits centred on privacy violations).

Instead of putting technology out on social media or selling to private citizens, Clearview targets the same clients as Israel’s NSO Group, which sells Pegasus: Clearview sells only to state agencies and law enforcement. Even stringent privacy protection laws such as the EU’s General Data Protection Regulation allow privacy violation to solve crimes and prevent terrorism. Mr Ton-That claims he never sells to authoritarian regimes. China’s Sensethink Technologies (which is rated to be as accurate as Clearview) occupies that space. Clearview has been extensively used by the Indian police since early 2020 to analyse pictures shot during protests, and collected off street CCTV networks.

Media revelations led to outcry and lawsuits from privacy advocates. Australia banned use of face-recognition. Several US police departments stopped using it, and legislators started looking at implications. Ms Hill winds up with a review of anti-face-recognition measures, which range from legislation, to funny masks, online pixel manipulation, wearing clothes with multiple faces printed, and taking data-sets for training offline.

China uses face-recognition to keep citizens compliant. India looks to be heading that way. Ironically for a Republican-funded company, Clearview was used to identify January 6 rioters. Mr Ton-That also offered Clearview free to Ukraine to help identify war casualties. The Ukrainians used it to send taunting messages on social media to families of dead Russian soldiers.

This particular genie is out of the box. Clearview is by no means the only operator in this space. Kudos to Ms Hill and others who have, at the least, revealed the dystopian rot. This book is good, competently written reportage which offers a useful, if overtly American perspective.