The Unpublished / Issue 001
The Machinery of Silence
A critical essay drawing on published evidence and documentary analysis.
Censorship in the Age of Algorithmic Surveillance
How ownership, economic pressure, intimidation, and data-driven systems can restrict information without formally banning it.
Read the paperImagine a newspaper that is legally free to publish an investigation but financially unable to survive the consequences. Its reporters have evidence. Its editor has questions about legal fees, advertising contracts, and next month’s salaries. The story remains unpublished.
Nothing in this hypothetical situation requires an official censor. Nobody needs to confiscate a printing press or issue a public prohibition. The pressure enters through the conditions that make publication possible.
This is the problem with treating censorship only as the removal of words. It directs attention toward the moment someone orders a story to disappear, while leaving other questions unresolved. Who could afford to investigate it? Who felt safe enough to become a source? What happened between publication and the article reaching an audience? Could someone read it without worrying about the record that reading might leave?
These questions do not describe a single coordinated system. A government monitoring its critics, a proprietor protecting a commercial interest, and a platform ranking content are different actors performing different actions. Their motives, responsibilities, and consequences should not be collapsed into one explanation.
Yet examining them together reveals a larger problem: freedom of expression depends on more than permission to speak. It also depends on the conditions under which information can be produced, circulated, and sought.
The machinery of silence operates wherever those conditions become restrictive enough that an explicit ban is unnecessary.
Before the story exists
A publication’s independence cannot be judged only by whether the state owns it. Private ownership can also create conflicts between journalism’s public responsibilities and the interests of those financing it. UNESCO’s 2020 study of media capture examined how political and economic actors can undermine independence through ownership, regulatory control, and the allocation of public advertising. Its concern was not simply biased wording within individual articles, but control over the institutions that decide what becomes news. UNESCO
Consider the difference between correcting a poorly supported allegation and refusing to investigate a powerful advertiser. Both might result in a story not appearing. But they are not equivalent editorial decisions. The first protects the integrity of reporting; the second, in this hypothetical example, protects a financial relationship from scrutiny.
This distinction matters because omission alone does not prove censorship. Editors must select, verify, postpone, and sometimes reject material. A serious criticism of media power must ask what shaped a decision, whose interests it served, and whether journalists could challenge it without retaliation.
Economic vulnerability creates another problem. UNESCO’s 2021/2022 global report described a journalism industry weakened by declining advertising income, newsroom closures, and job losses, with the pandemic intensifying those pressures. Supporting the financial viability of journalism, the organization argued, must go together with protecting editorial autonomy. UNESCO
The implications extend beyond the survival of individual newspapers. Imagine two potential stories: one can be assembled quickly from a press release; the other requires weeks of interviews, travel, document analysis, and legal review. A depleted newsroom might publish the first and abandon the second without anyone ordering it to protect a particular institution.
That is not automatically deliberate censorship. It is a loss of investigative capacity. But the distinction should not make the public consequence unimportant. A society can lose access to essential reporting through financial exhaustion as well as political interference.
The question becomes not only whether journalists are allowed to investigate power, but whether they have the resources to do so.
The cost of speaking
Pressure does not always have to prevent publication directly. It can make speaking so costly that withdrawal becomes a reasonable act of self-protection.
Strategic lawsuits against public participation, commonly called SLAPPs, illustrate this mechanism. In its 2024 recommendation, the Council of Europe described legal actions threatened or pursued to intimidate people and restrict expression on matters of public interest. It also recognized how prolonged proceedings and financial burdens can discourage public participation. Council of Europe
The threat in such a situation is larger than the possibility of losing a case. A small publication might be confident in its reporting and still lack the money to defend it through years of litigation. Legal vindication at the end would not necessarily restore the time, income, or reporting opportunities lost along the way.
Not every lawsuit against a journalist is abusive. People must be able to challenge genuinely harmful falsehoods. The relevant distinction is between seeking a remedy for an injury and using the legal process itself to punish scrutiny.
Harassment can exert pressure through a different route. In May 2026, UNESCO summarized research in which 45 percent of surveyed women journalists and media workers reported self-censoring on social media because of online violence, while 22 percent reported self-censorship at work. These figures describe the survey respondents, not a population-wide estimate for all women journalists. They nevertheless document how abuse can affect what participants feel able to say. UNESCO
Calling this self-censorship can make the decision sound entirely internal, as though a journalist simply became less willing to speak. The term should not obscure the external pressures producing that decision.
There is an ethical danger in demanding courage from people while treating the conditions threatening them as inevitable. A journalist who withdraws from a subject after sustained intimidation has not necessarily abandoned professional responsibility. The institutions meant to support that journalist may have failed to make responsible reporting survivable.
The useful question is not why someone lacked the courage to continue. It is why continuing required such a disproportionate personal sacrifice.
From being watched to being classified
Surveillance introduces another kind of pressure: the possibility that an action will be recorded, connected to other information, and interpreted later.
In this essay, algorithmic surveillance means the collection and automated analysis of information about people to identify patterns, infer characteristics, or categorize behavior. It is more than someone watching a screen. The consequential act may be the conversion of scattered records into a profile that can be searched, compared, or used to make decisions.
The United States Federal Trade Commission’s September 2024 examination of major social media and video-streaming services documented extensive collection and retention of personal information. It also found that companies fed information about users and non-users into automated systems, while people had little or no ability to opt out of some of those uses. The report connected these practices to business incentives, especially targeted advertising. Federal Trade Commission
A record of behavior and an interpretation of that behavior are not the same thing. Visiting a location does not necessarily establish someone’s beliefs. Reading about a political movement does not demonstrate membership. An inference can be consequential without being accurate.
The FTC’s action against data broker Mobilewalla provides a concrete example of how commercial data collection can intersect with political activity. In December 2024, the agency alleged that the company collected sensitive location information from advertising exchanges and other sources. It also described a June 2020 report in which Mobilewalla analyzed people protesting the death of George Floyd and inferred characteristics including racial background and whether they lived in the cities where they protested. Federal Trade Commission
The FTC finalized an order restricting Mobilewalla’s collection and sale of sensitive location data in January 2025. Federal Trade Commission
This case does not establish that advertising systems and state intelligence operations are interchangeable. Nor does it show that every commercial profile is used for political repression. It demonstrates something more specific: information gathered through commercial infrastructure can be repurposed to categorize participation in public life.
The concern is therefore not only secrecy. It is also the authority given to interpretation. Who decides what a record means? How can someone contest an inaccurate inference they do not know exists? What prevents information collected for one purpose from becoming a tool for a more intrusive one?
A system does not have to understand a person to make that person vulnerable to classification.
The story ends at the source
For journalism, surveillance threatens more than a reporter’s private life. It can undermine the relationships through which reporting becomes possible.
A February 2024 investigation led by Access Now, with Citizen Lab and other partners, documented at least 35 people in Jordan who had been targeted or infected with Pegasus spyware. The cases covered a period from at least 2019 to September 2023 and included journalists, lawyers, activists, and other civil society members. Citizen Lab distinguished evidence of targeting from confirmed infection, an important limit when interpreting the findings. Access Now
Targeted spyware is not the same mechanism as algorithmic profiling or content ranking. It belongs in this discussion because it demonstrates another way surveillance can compromise the conditions necessary for expression.
A source considering whether to share evidence must evaluate more than the trustworthiness of a journalist. They must also consider whether confidentiality can be maintained. UNESCO’s study Protecting Journalism Sources in the Digital Age identified mass and targeted surveillance, data retention, and overreaching security measures as threats to source protection. It warned that weakened confidentiality can discourage whistleblowing and undermine public access to information. UNESCO
The public interest in privacy becomes clear here. A confidential conversation may be the beginning of an investigation into wrongdoing. Protecting that conversation is not necessarily about keeping public affairs secret. It may be what makes their eventual disclosure possible.
This reverses the familiar suggestion that privacy and public knowledge are opposing values. Under these conditions, privacy is part of the infrastructure of public knowledge.
If a source reasonably believes that approaching a journalist could expose them to retaliation, the loss occurs before publication. There is no finished article to remove. There may not even be a recorded interview.
When the reader starts self-editing
The pressure can extend to people seeking information, not only those producing it.
In a 2016 study, legal scholar Jonathon Penney examined traffic to privacy-sensitive Wikipedia articles following the June 2013 revelations about NSA surveillance. He found an immediate decline and a subsequent change in traffic trends consistent with a chilling effect. Because the research analyzed aggregate behavior, it could not establish what motivated every individual reader or prove that all changes resulted from surveillance concerns. Schulich Law Scholars
Its significance lies in the possibility it identifies: information can remain accessible while people become less willing to seek it.
A person does not need to believe that every search is being inspected. Uncertainty about whether an inquiry might be misunderstood can be enough to make avoidance seem safer. The resulting constraint concerns the freedom to investigate unfamiliar ideas, including ideas one might ultimately reject.
For that reason, the absence of blocked websites is an incomplete measure of intellectual freedom. A stronger test asks whether people can investigate a difficult subject without having to calculate what their curiosity might later be taken to mean.
Published is not the same as reachable
Even when a story is investigated, written, and published, its circulation raises a separate set of questions.
Platforms do not simply display everything in a neutral sequence. In its public explanation of Facebook and Instagram ranking, Meta described systems that predict the value of content to a user, including predictions about whether someone might share a post. It also distinguished ranking from measures that remove harmful content or reduce the distribution of other material. About Facebook
These mechanisms need to be separated carefully. Surveillance concerns the gathering and analysis of information about people. Moderation concerns the enforcement of content rules. Ranking concerns the order and prominence of material. They can interact, but evidence about one does not automatically prove wrongdoing in another.
Nor is every low-performing post evidence of suppression. An article may attract little attention for reasons unrelated to censorship. Some form of selection is unavoidable in a feed containing more material than anyone can read.
The important question is how consequential selection decisions can be examined and challenged.
A useful example comes from BSR’s 2022 human rights review, commissioned by Meta, of the company’s actions during the May 2021 crisis in Israel and Palestine. The review identified adverse impacts on Palestinian users’ expression, erroneous removals and account penalties, and failures to remove some harmful content. It did not identify intentional bias, but did identify unintentional bias, including problems involving the handling of Arabic content and regional or dialect expertise. bsr.org
Meta acknowledged concerns about unintended impacts on Palestinian and Arab communities and announced commitments responding to the review. It also emphasized the need to prevent its services from facilitating hatred or violence. About Facebook
The distinction between intent and outcome matters. A harmful result does not establish that an employee deliberately sought to silence a community. But the absence of demonstrated intent does not make the result harmless or remove the responsibility to correct it.
This is why the argument should not be that all moderation is censorship or that platforms should circulate everything without restriction. Threats and harassment can themselves make participation unsafe. The stronger demand is for rules that are understandable, enforcement that accounts for linguistic and political context, and meaningful remedies when decisions are wrong.
An explanation that “the system made a mistake” should begin an investigation, not end one.
How separate pressures can reinforce each other
The mechanisms described here should be understood as potentially reinforcing pressures, not evidence of one hidden command center.
Consider a hypothetical local investigation. A newsroom has little money for extended reporting. A potential source fears identification. The publisher anticipates expensive litigation. After publication, the article depends heavily on a platform whose distribution decisions are difficult for the newsroom to interpret.
No single event necessarily explains the final result. The story might be shortened, delayed, abandoned, or published without reaching the people most affected by it. Each stage presents a different problem and requires different evidence.
This is also why the phrase machinery of silence should not become an excuse for vague suspicion. Claims of interference need documentation: ownership arrangements, advertising decisions, legal correspondence, forensic findings, moderation notices, or research into distribution patterns. A dramatic interpretation is not a substitute for establishing what happened.
The common concern is the gap between formal permission and practical possibility.
A publication might possess the right to investigate while lacking the resources to sustain an investigation. A source might possess the right to speak while facing a credible threat of retaliation. A reader might possess access to information while feeling unable to seek it safely.
Treating each situation as a purely individual choice would miss the conditions shaping those choices.
What freedom would require
An adequate response cannot stop at telling journalists to be braver, readers to become more skeptical, or users to adjust their privacy settings. Those actions may have value, but the problems described here also concern institutions.
For journalism, the priority should be independence that survives financial pressure: transparent ownership, clear separation between commercial and editorial decisions, and funding arrangements that do not give a single patron control over what can be investigated. Legal support and effective safeguards against abusive proceedings should make accountability reporting less vulnerable to exhaustion. UNESCO’s work on media viability and the Council of Europe’s anti-SLAPP recommendations support these institutional approaches. UNESCO
For data systems, the central question should be whether information needs to be collected at all, not merely whether a lengthy policy discloses its collection. The FTC’s 2024 report recommended limits on collection, retention, and sharing, alongside stronger user control and more rigorous oversight of automated systems. Such measures address the accumulation of power before it has to be challenged after a harmful use. Federal Trade Commission
For journalism’s sources, confidentiality should be treated as a public-interest safeguard. Independent oversight of surveillance and meaningful source protections matter because the consequences of exposure extend beyond the person whose privacy is violated. They can affect what everyone else is able to learn. UNESCO
For platforms, accountability should mean more than publishing general descriptions of their systems. People need intelligible reasons for consequential restrictions, usable appeals, and scrutiny capable of identifying patterns across languages and communities. The aim should be neither unlimited corporate discretion nor government control over legitimate disagreement.
None of these measures guarantees a perfect information environment. They would, however, distribute the burden more fairly. The person reporting wrongdoing should not be left to solve newsroom economics, abusive litigation, surveillance, and opaque platform decisions alone.
The final test is not whether every thought finds an audience. It is whether people have a meaningful opportunity to investigate, communicate, and question without being priced out, threatened out, or monitored into withdrawal.
A story can be lost before anyone writes it. A source can fall silent before an interview begins. A reader can turn away before opening a page.
A serious defense of expression must account for those absences, not only the words that were visibly crossed out.
References
Sources cited in this essay, collected from its inline references.