Surveilling the Surveillor: Assembling a Digital Dossier of Hoan Ton-That
In the past year or so, a few media observers have shone a spotlight on Clearview AI, a tech company that facial recognition tool that extracted data from, as their website describes, “public-only web sources, including news media, mugshot websites, public social media, and many other open sources.” The company is infamous not only for its ethically-questionable practices of data harvesting and breaches of privacy, but also for its ties to law enforcement and alt-right political figures and groups. A recent article in the New York Times traced this connection by probing into the people who helped its founder and CEO Hoan Ton-That create this tool, specifically, Republican politician Richard J. Schwartz (the co-founder of Clearview AI), the alt-right media figure Charles Johnson, and Silicon Valley CEO Peter Thiel.
However, its analysis of Hoan himself is less clear. This article in particular does not probe into Hoan’s political beliefs, but rather seeks to indict his company through his social network alone. Other articles also hardly give a picture of Hoan’s personal beliefs or identity, and at most just describe the strange series of apps he created (from “Trump Hair” to various phishing scams). Perhaps for other authors, his social network and the technology he created are enough to prove some moral guilt, but, as I would soon discover, they missed so much more direct evidence.
Nevertheless, these authors’ shallow description of Hoan might in part be caused by the general lack of surface-level information about him. A cursory Google search shows his personal website, his Wikipedia article, and various news articles about his role in Clearview AI. On YouTube, one can find various interviews questioning him on his company, and perhaps more notably, a personal YouTube account with five year old videos of him playing the guitar.
Most other people only care about Hoan for his role as creator of Clearview AI. Personally, I Googled him because I thought it was interesting that a Vietnamese person was the founder, but what drew me in was his appearance: the contrast between his younger pictures and the older, awkward-looking images from more recent interviews, a contrast wide enough that I wonder if his own software can identify them as the same person. The recent pictures fit my image of what a Silicon Valley tech CEO would look like: a boring suit, receding hairline, nerdy wire-frame glasses. But his younger pictures showed someone entirely different: a fashionable, long-haired androgynous person; in other words, someone I found myself identifying with. And this discrepancy between these images — the supposedly evil, socially-disconnected tech CEO that the articles describe versus this androgynous model—got me desperate to know more.
And so I clicked on his personal website, and at the bottom I saw a short line that deepened my curiosity even further: “Pronouns: he/him.” Someone with ties to alt-right groups was not the type of person who I thought would list their pronouns. That, combined with his appearance, raised the question in me: what if he was nonbinary? And so, to see if there was any context for these pronouns, or perhaps if they were different in the past, I checked on archive.org’s Wayback Machine (WBM) for any results on his webpage. However, when I saw how far back in time the archive went, my curiosity grew beyond that initial question, and I soon found myself in a rabbit hole of online archives. The process of searching through these archives was eclectic and often started from the future and went backwards, but for the sake of comprehensibility, the following section will trace Hoan’s online history in a chronological order.
The Biography of Hoan Ton-That Online
Hoan Ton-That is a Vietnamese-Australian born in 1988. His first presence on the internet stored in the WBM was in 2004 at http://www.ton-that.org/, a personal website different from his current one. The first capture, from September, 23 2004, only has a brief message, “I am hibernating (ie, going to school). I will be back in the school holidays! Happy Hacking.” In 2004, Hoan was 16, which means this school is likely high school. On the January 26, 2005 capture, we can see a directory that shows, for one, that the website dates back to at least July 16, 2004, and furthermore, that Hoan had added various other sections to his website, such as a blog, plans, and a wiki. Most of these branches were not archived, except for the journal. However the journal only has a file that sets the formatting, not any content.
The next major update to the archive is two years later on February 17, 2007, which now shows a list of apps that Hoan has worked on since 2006. The website would be updated and captured a few more times in the following months, and each capture has functioning subbranches that I have yet to fully dig through. Important from this section is the now broken contact email hoan@ton-that.org, as well as a dating app that he made, Touchfelt, which reveals that he might have gone to ANU, or the Australian National University.
However, 2007 was also the year in which he dropped out and moved to San Francisco. Sadly, the captures after June 2007, although fairly frequent and going all the way up to the end of 2008, seem to be broken. After this, the archive of this site would have nothing else until 2011, with a series of automated crawls of a dead url. Thus, we can assume that he abandoned this website sometime after 2008 but before 2011.
Hoan’s next online appearance would be on Twitter, with his Twitter handle simply being @ Hoan (now defunct). The earliest archive of his account is from February 2008, an unfamiliarly colorful and customizable era of the internet. This first capture includes a link to his first personal website, a few of his early followers (many of whom are still active and quite famous on twitter) and a simple bio, “I’m just a kid!” However, his tweets (or “updates” as they were called back then) were protected, and so we never get to see any tweets directly from Hoan himself.
Nevertheless, the changes to his bio are revealing enough, as in the February 2009 capture, we see a new bio, “Anarcho-Transexual Afro-Chicano American Feminist Studies Major.” A contemporaneous article from a proto-alt-right media source would take this description at face value and disparage Hoan as being just another Silicon Valley queer liberal-leftist. However, it is blatantly obvious that this bio is ironic, and meant as an insult to trans people rather than being a description of himself (since he is very clearly neither black nor Mexican either).
After 2009, the archive of this twitter account stops. This same year, however, would have the first iteration of Hoan’s current personal website. This capture has a brief description of the apps Hoan has coded, as well as the same transphobic Twitter bio at the top. The description also shows that he used this site as a blog, although none of that has been preserved.
This capture is the only one in 2009, and the site will not be saved again until December 2013. But before jumping there, I will detour to Hoan’s first explicit appearance on YouTube: a video of him playing guitar in Seattle on September 10, 2009.
The description of this video describes a “Chantelle Tibbs,” and a Google search for this person reveals that Hoan and Chantelle were in fact married (but divorced in 2014). Nevertheless, there are a variety of results showing them together, including Hoan’s actual first appearance on YouTube a few months earlier in April 2009, an interview with the two containing a brief clip of Hoan’s voice, and Chantelle’s undeleted twitter account with a tweet that mentions Hoan’s old twitter.
Returning to Hoan’s personal website, the December 2013 capture is short but gives links to another email address (hoan.tonthat@gmail.com), his twitter account (which is how I discovered it in the first place), and a link to a website called AngelList (I have still not yet devled into this branch of Hoan’s history).
The next major update to Hoan’s website is in August 2015, which has a full personal description and mostly functioning formatting (except the image slideshow, which only works in the October capture). This iteration has links to several social media accounts, including his AngelList, Facebook (since deleted), Instagram (images are broken), Tumblr (unchanged and still operational today!) contains just a few gifs of himself made with Shimmer, an app that he himself made), and most importantly, his second Twitter account, @hoantonthat.
The WBM only has two captures of this twitter account, but the tweets that it saved are illustrative enough of his political views. The first capture from August 21, 2015 does not have too much. He supports range voting, opposes the Euro, and looks up to monetarist economist Milton Friedman. However, interestingly, last tweet in this capture references the GOP primary debate on August 6th, i.e., the first debate that Donald Trump appeared in.
The Rorschach test that he is referring to is a psychological test that analyzes how someone perceives inkblots in order to determine their mental state and personality. I myself am still unclear as to exactly what he means by this. Perhaps he sees the Republican candidates as the inkblots, and how someone perceives the candidates is a way to determine that person’s political personality. Or perhaps the candidates are the test subjects who are being shown the inkblots, and this might be a comment on Trump’s mental state. Regardless, this comment still feels too vague to give a hint at Hoan’s political stance regarding Trump.
However, the second capture from January 17, 2016 is far more illustrative. He frequently retweets alt-right figures and media sources, such as infamous gay right-wing media figure Milo Yiannopoulos or links to Breitbart. He himself tweets, “In today’s world, the ability to handle a public shaming / witch hunt is going to be a very important skill,” which is his attempt to jab at cancel culture (and perhaps also his consent to what I am doing right now). Furthermore, he also complains about how big cities in the US are too “liberal.”
Although these are only two small snapshots of what could have been a fairly active twitter account (based on the frequency of tweets in the second capture), what we little we have is probably enough to determine his own personal alignment with the alt-right. Even if he did not identify himself as such, he clearly shares many of the same beliefs. Of course, it has been five years since these tweets and so his views might very well have changed, but because he no longer has an active presence on twitter (except this one dead and unnoticed account, which as I will explain below, is likely authentic), I have nothing else to go on. Furthermore, I am suspicious of the things he says in more recent interviews, because as CEO of Clearview, his main priority is the marketability of both himself and his product. Thus, even though the recent New York Times article describes Hoan criticizing the January raid of the Capitol, this is probably just a way to position himself on the side of law-and-order (his main constituency) rather than a disavowal of Trumpism.
Returning to Hoan’s personal website, there would be a few edits during 2016. Namely, the slideshow was replaced by a single picture of him holding an umbrella, the link to his tumblr was removed, and notably, the personal statement, “I also love to play soccer and guitar, and collecting clothes! I’m now traveling the world and working on the next big thing. Remain excellent,” is deleted as well. This leaves his description as only containing information about his professional history, but in 2017 even this is removed, leaving just the photograph and his social media links.
In general, these changes feel like a depersonalization of his public image, an attempt to conceal information about himself. The first few changes reflect a desire for greater marketability, removing the personal and presenting only professional achievements. However, the complete removal of this professional history occurs the same year that Clearview AI was founded, and so he might have buried these achievements because they might get perceived as inferior and harm Clearview’s perception.
The website would remain unchanged until January 17, 2020, a day before the publication of the New York Times article that shot him into the spotlight. This new iteration would set the formatting that is used today, although there are some key changes that occur between then and now. The most noticeable change in this new iteration is the new picture: somewhat more formal than the umbrella picture, but still relatively fashionable and young. The description is similar to the 2016 one, mostly describing his professional achievements, but now it includes his pronouns at the end.
On January 20th, he puts Clearview AI in his introduction, perhaps in response to the media attention around this time. But on the 23rd, he adds a detail that stands out from his entire media presence hitherto:
I previously went by other gender identities but currently go by: he/him.
This statement is the first and only authentic self-description of his prior exploration of gender that I was able to find in the archives. The only other source I found regarding Hoan’s gender was a news article describing him as genderfluid, but I have not found anything from Hoan himself to corroborate that. There was, of course, the apparently sarcastic twitter bio from 2009, but in light of this recent authentic statement, perhaps that bio was not a straightforward irony. Although it is very clearly in mockery of trans people, it might have been meta-ironic: using the ironic construction to simultaneously express transphobia while also hiding a grain of truth in a way that evades scrutiny from his peers.
But the content of this January 23rd edit is not the only detail that stands out. In the rest of the description, he describes himself in the third-person, but for his gender identity he suddenly shifts into the first-person. This shift is a rupture in the professional, marketable, and impersonal identity that he built up on the internet. It is this use of “I” that makes this gender description feel so genuine, as opposed to the transphobic sarcasm of the 2009 twitter bio.
However, this brief bit of authenticity would not last, as Hoan would gradually remold his website and his image overall to match the cis-binary-masculine expectations of the tech and business world. On May 9, he would change his picture from the youthful and more casually-dressed one to a face-on portrait of him in a suit. Although his hair is still long, he presents it in a more masculine way in this new picture. This in general matches the image he cultivated for himself when he was interviewed in the media, where he would always be wearing a suit.
The next step towards professional marketability is on September 28, when he overhauls his description entirely. For one, he removes any mention of his past gender identities, leaving just “Pronouns: he/him.” Looking to the grammar of this bio, not only did he remove the single “I” statement on his profile, but he even changes how he refers to himself in the third-person. In the previous description, he always used his first name “Hoan,” but in the new one, he instead uses his full name at the start and “Mr. Ton-That” for the rest.
There are a few more WBM captures after September 28th (the last one, April 16th, being done by me), but there would be no further edits. This September 28th edit is thus the last glimpse into Hoan’s personal motivations. He does appear in interviews with the media after this point, but I have not deeply dug through these yet, and even then, the interviewers likely care far more about his company than about Hoan himself. Outside of his website, Hoan has no active public social media presence. I have only been able to find a twitter account, @HoanTonT, which has no profile picture, no tweets, no likes, just 18 followers, and a two-word bio, “Coming Soon.” The profile was made in January 2020, so evidently this soon is not to come.
There is, of course, the possibility that someone caught on the hype around Hoan and Clearview AI and made an imposter account. However, if this website is correct, this account was made on January 15th, which is still a few days before the New York Times article was published. And if Hoan was renovating his website at around this time, it makes sense that he would at least intend to update his twitter as well.
Yet in the end, he never did become active on twitter or anywhere else. Perhaps this might be because of his disputes with the tech giants over his use of the data they had harvested themselves, but it might just be intentional. He might be a shrewd businessman and understands that any superfluous popular attention on his company might be bad for its survival, and so he lays low. Perhaps he wants to ensure that his product will be marketable in all political climates, and so he avoids making public political statements backfire on his attempts to curry favor with law enforcements and local governments. Or perhaps his experience in the creation of Clearview taught him how precarious his own privacy really was (it may be no coincidence that the last archives of his Twitter and Instagram come in 2016, a year before the founding of Clearview). And so, with his privileged access to power, Hoan’s absence from social media might just be an attempt to protect the privacy that he has denied everyone else.
A Personal Exercise of/in Surveillance
The project I have set out on, to dig up as much information online about Hoan and construct a history of his person, can perhaps be seen as an attempt at retaliatory surveillance. However, our forms of surveillance are not equivalent. Hoan’s Clearview AI embodies the surveillance of the digital capitalist age: harvesting vast amounts of data, or behavioral surplus as Zuboff describes, in order to produce and refine a facial recognition software that gets sold to law enforcement. This digital surveillance can process data from hundreds of millions of people, but for any given person, the intrusion feels minimal. Digital surveillance does not probe deeper than what is at the very surface level, and that is because it does not need to go any deeper. Using Deleuze’s terms, the fundamental unit of digital surveillance is not the individual but the dividual. It only cares for certain aspects of a population, and does not care for the true identity or beliefs of any one person.
However, my practice of surveillance was the exact opposite. My very end goal was to figure out, or perhaps to construct, the individual Hoan Ton-That. Discovering minor details, like a music meme that he liked in 2009 , felt like a major breakthrough because I felt that any of these seemingly irrelevant details might be useful in getting a glimpse of who Hoan really is. Hoan’s surveillance involves minor intrusions multiplied across the entire population, while my surveillance is a deep, highly intrusive probe into just one person’s personal history. Perhaps Hoan would not mind, but for me, if I learned that there was someone else who was scrutinizing me as much as I had scrutinized Hoan, I would be deeply unsettled, far more unsettled than by the surface-level intrusion that Clearview carries out on me.
Our surveillances are asymmetric due to the differences not just in our goals but also in our capacities: I do not have access to programs that can scour the internet, but I do have a brain. A brain that can make inferences, that can read between the lines in several year old statements in order to figure out the “true” meaning behind them. Meanwhile, I am no AI expert, but I presume that machine learning systems at today’s level cannot infer and interpret statements like a human can (yet). And to those performing this digital surveillance, this level of interpretation is unnecessary because they care about profit, not some hidden truth. These data-gathering systems only care about more data, and quantity alone is sufficient in producing a profitable product.
Thus, the fact that my end goal is a “true meaning” or “true self” puts me closer to the political surveillance of 20th century authoritarian states. Although my medium of surveillance is digital archives and social media, my capabilities and methods are more analog: I need to dig up and find information, sift through all the data, and make inferences all on my own. What I do on an individual scale is what socialist states had to do on a population-wide scale, and it is no wonder that states such as East Germany or Romania required so much manpower, in the form of a surveillance bureaucracy and informant networks, in order to perform that level of in-depth analysis on anyone who might be deemed suspicious.
Both my surveillance and the political surveillance of these ideologically-motivated states believe that some truth exists, and that the aim of surveillance is to find information that can reveal such truth. Of course, in many cases, the “truth” that was believed to exist were just the surveillor’s preconceived notions, and the information meant to act as evidence was stretched or outright fabricated to prove those preconceived notions. And in a way, my own surveillance does the same. I of course could not extract information through an interrogation (sadly), but I did go into the archive of Hoan’s online self with a hermeneutics of suspicion. I had a preconceived expectation that Hoan was guilty (i.e. right-wing) in some way, and the aim of my surveillance was to find information to prove his political guilt.
Of course, I cared a lot about the small, personal, seemingly apolitical details, as did state surveillors. The purpose of these small details is to refine and construct a better understanding of the individual, such that these details can be placed in reference to the political guilt. And in the case of Hoan, I needed to gather these details in order to somehow reconcile how he was both an alt-rightist and formerly nonbinary, and so I saw the random things he coded in the past or the videos of him playing the guitar as a “glue” that can somehow piece together the individual of Hoan Ton-That.
But even though I analyzed Hoan with this hermeneutics of suspicion, the process of looking so deeply into these minor details personalized Hoan in my mind. As time went on, Hoan had increasingly depersonalized and professionalized his online presence, but my inquiry backwards in time showed me what really did feel like an authentic Hoan. Even though the truth I was seeking was in part a proof of political guilt, the apolitical side of the authenticity I unearthed was still enough for me to form an attachment, to sympathize with my object of surveillance just like the Stasi agent in The Lives of Others.
Even as I searched for evidence of his ideological evils, another part of me hoped that he could “change for the better.” And, indeed, if I were a state, the end of my information gathering would not just stop at constructing an individual but would seek to use this information to effect that change, to mold him into someone who doesn’t align with the alt-right. Nevertheless, as Foucault has infamously argued, such discipline cannot occur without a constant visibility. And if Hoan refuses to make himself visible and attempts to opt out of the surveillance system that he has created, then any change, whether in Clearview AI or in Hoan himself, must come about through this bottom-up political sousveillance.
