“Algorithmic outing” sounds like an abstract concept, almost science fiction. It is not. It is a very concrete mechanism, already documented repeatedly for over a decade, and I believe that in our community we should understand exactly how it works, because only then can we demand that it be corrected.
The mechanism: how a social network “guesses” who you are, even if you haven’t told it.
It all starts with a function that almost no one questions: “People you may know.” Facebook acknowledges that it uses more than 100 different signals to generate it, mutual contacts, shared networks or groups, contact lists uploaded by other people, but it refuses to detail which ones exactly come into play in each case, as investigative journalist Kashmir Hill revealed in Gizmodo. That point is key: the platform itself admits that it uses data that goes far beyond what the user has intentionally shared, and it does not explain how.
The consequences of that opacity have already been documented several times. In 2012, a bigamous man was discovered when Facebook sent mutual friend suggestions to his two wives. In 2016, the social network even recommended the patients of a psychiatrist to befriend each other, probably because they had all shared the waiting room with their phones on, inadvertently exposing that they shared a therapist. And in 2017, several sex workers discovered with horror that Facebook suggested their clients as friends on their personal profiles, and their personal profiles as friends to their clients, deliberately breaking a separation between identities that they had built with different accounts, emails, and phones, as Kashmir Hill also documented for Gizmodo. None of those people did anything to make it happen. The algorithm simply connected dots that they had tried to keep separate.
Why this happens: it is not a flaw, it is the design.
Here is the point that interests me the most to highlight, because it completely changes how the problem should be read. This is not a technical error that someone has not fixed yet: it is the direct consequence of a business model built on “connecting” people as much as possible, without friction and without secrets, as a design principle. The more connections a social network detects and suggests, the more time people spend on it and the more data it generates. Privacy, in that design, is not a goal: it is, at best, an optional adjustment that needs to be sought out, and at worst, an obstacle to the platform’s growth.
For most users, that aggressive design of “connecting everything with everything” is just a little annoying. For someone who deliberately keeps a part of their identity separate, an LGTBIQ+ person who has not come out in their work or family environment, a trans person in transition, someone participating in community spaces they prefer to keep private, that same design can be, literally, dangerous.
The most extreme variant: when the algorithm becomes a data market.
The most documented case of this last idea has a proper name. In July 2021, Jeffrey Burrill, Secretary General of the United States Conference of Catholic Bishops, the institution that coordinates the official stance of the Catholic Church in that country, including its doctrine on homosexuality, resigned immediately after the publication of a report by The Pillar. He had not been hacked: the media legally purchased from a commercial data intermediary, a history of signals from the Grindr app linked to an identifier of his mobile phone, and cross-referenced it with the location of his home and office for almost three years, between 2018 and 2020, until identifying him with almost daily precision, as confirmed by The Pillar itself and later reported by Axios, NBC News, and The Washington Post.
Here it is worth pausing on the uncomfortable nuance, because it is the one that truly generates debate: Burrill did not legislate against LGTBIQ+ people nor did he actively promote any persecution; his “fault,” according to the conference’s own statement, was not living with integrity his vow of celibacy, not attacking anyone from his position. It is tempting to celebrate a little the fall of someone who represented an institution that harms our community while he himself secretly lived a gay life. But if we accept that it was right to expose him for that contradiction, we are also accepting that any data market, without any judicial or ethical control, can decide tomorrow who “gets to” be exposed and that power, applied in the opposite direction, is exactly what has historically been used to persecute us. The method of buying “anonymized” signals and cross-referencing them with location does not distinguish between a hypocritical ecclesiastical position and any LGTBIQ+ person with nothing to lose.
The most recent variant: when you don’t even have to decide anything.
The mechanism is not limited to data markets with interested buyers. A study published in July 2026, with in-depth interviews with 20 LGTBIQ+ people aged 18 to 60, documents a quieter and more everyday version of the same problem: content recommendation algorithms that show, on someone’s mobile, posts or ads clearly identifiable as LGTBIQ+ while that person is on the bus, in a café, or at work, in what researchers call “hybrid spaces,” without the person having posted or searched for anything at that specific moment. Many of the interviewed individuals admitted to using separate accounts or constantly monitoring what appears on their screen in public, precisely to manage a risk that does not depend on what they do, but on what the algorithm decides to show, as reported by the research disseminated by Phys.org.
The counterargument that must be acknowledged.
It is fair to acknowledge that platforms have responded, albeit late and partially: there are options to limit who sees your friends list, to partially deactivate “People you may know,” to request that your contacts not be uploaded. The problem is that these options function as patches over a design that is still built, by default, to maximize detected connections, not to protect those that someone has decided to keep separate. Having the option to protect yourself is not the same as being protected by default, and the difference between the two is precisely what the cases of the psychiatrist, the sex workers, Burrill, and the 2026 study documented.
What is really needed.
We need to demand real transparency from platforms about what signals they use to “connect” profiles, not a generic list of more than 100 factors without detail, but concrete explanations when something goes wrong, and we need the default option to be privacy, not maximum connection. We also need, as a community, to resist the temptation to applaud these methods when the exposed person is someone we dislike: if the data market that brought down Burrill seems acceptable to us, we will have no argument when that same market is used against any of us. As long as the design continues to reward detecting links over respecting the boundaries that someone has set between their identities, algorithmic outing will continue to occur, not as an exception, but as an expected result of the system itself.
