The CIA once needed hundreds of journalists to shape what people read. Today a handful of people with admin access shape what billions of people see. Same objective, better tools. And now governments are deciding who gets let through the door — starting with the children.
This file was first published on 5 March 2026 as part of the They Own The Sky series. Its argument is simple: every generation, a new medium; every new medium, the same hands reaching for the dial. First the press, then the platforms, and now the gate in front of the platforms.
On 23 October 2026 we rebuilt it to the site’s current standard. Every claim now links to its source. Several claims in the original did not survive that check, and we say so plainly in the Editor’s Note below. The core argument survives; some of the scaffolding didn’t.
The through-line: influence over what people see has moved from newsrooms to code, and the documented record of agencies leaning on the platforms is real. The Australian age ban is not, on the evidence, an identity database — but enforcing it is teaching platforms to profile everyone more closely to work out who is a child.
We went back through every claim in the March version against the primary record. Here is what changed. Where we got it wrong, we say so.
| Wrong | “Platforms must collect age-verification data on every Australian and hold it for a year.” The Act says the opposite: information collected for age checks must be destroyed after it’s used, and keeping it is a breach of the Privacy Act.9 eSafety says not every account holder has to be checked.11 We withdraw that claim. |
| Wrong | “They built the most comprehensive identity verification database in Australian history.” The ban creates no central database, and platforms may not force anyone to use government ID or Digital ID.9 The real privacy cost is different, and arguably worse in its own way: platforms now scan posts, captions and faces to guess who is a child (Section 04). |
| Sharpen | “Operation Mockingbird.” There are two Mockingbirds, and the March file blurred them. Project MOCKINGBIRD is in the CIA’s own files: a 1963 wiretap of two Washington columnists. Operation Mockingbird is the name Deborah Davis gave in 1979 to a CIA press-influence network. The CIA’s ties to journalists are confirmed by the Church Committee and Bernstein; what investigators never found is a single programme by that name run the way Davis described.326 (Section 01) |
| Sharpen | “400 journalists, including the publisher of the New York Times.” Bernstein counted 400-plus journalists, and separately named NYT publisher Arthur Hays Sulzberger among the executives who lent the CIA their cooperation.2 The Church Committee’s own count was about 50, in a report published in 1976, not 1975.1 |
| Sharpen | “2011: fake identities to flood social media with pro-establishment sentiment.” The contract was real, but US Central Command said the personas were for Arabic, Farsi, Urdu and Pashto sites, not US audiences or Facebook and Twitter.5 |
| Sharpen | “Twitter Files: a subsidiary of the FBI and CIA.” Matt Taibbi wrote that Twitter’s contact with the FBI was so constant it was “as if it were a subsidiary” — the FBI, not the CIA.6 The US Supreme Court later threw out the main lawsuit over this pressure because the plaintiffs couldn’t show it caused their own removals.7 |
| Sharpen | “Europe is moving in lockstep; von der Leyen called the law inspiring.” She said she was “inspired by Australia’s example.”17 Several countries are considering bans; most had not passed one as of March 2026.16 |
| Holds | The House vote of 102 to 13 and the eight days from introduction to passage;109 no parental-consent exception;9 the News Corp campaign behind the law;16 Frank Church’s 1975 warning;4 and former intelligence staff in platform trust-and-safety jobs (from a partisan but specific source).8 |
| Label | “They banned children because the algorithm stopped working on them.” This is our hypothesis, not a documented finding. We have kept it, clearly marked as inference (Section 03). |
New since March: the ban took effect on 10 December 2025; penalties have doubled; and on 20 October 2026 the Communications Minister said the government is preparing to sue five platforms; and from 1 December businesses — potentially including social media platforms — can start using myID (Section 04).
Start with what is not theory. In April 1976 the Senate’s Church Committee reported that, until a policy change two months earlier, the CIA kept covert relationships with about 50 American journalists or employees of US media organisations — about half of them paid.1 The same report found the CIA had produced, subsidised or sponsored well over a thousand books before the end of 1967.1
A year later, Carl Bernstein went further in Rolling Stone. Citing documents at CIA headquarters, he reported that more than 400 American journalists had secretly carried out CIA assignments over 25 years, and that executives including William Paley of CBS, Henry Luce of Time and Arthur Hays Sulzberger of the New York Times had lent the agency their cooperation.2
“Operation Mockingbird” is how most people know this history. There are actually two Mockingbirds, and both are real — as different things.
Project MOCKINGBIRD (1963) is in the CIA’s own “Family Jewels” files: a wiretap on syndicated columnists Robert Allen and Paul Scott from 12 March to 15 June 1963, after they published classified material. That was surveillance of journalists.326
Operation Mockingbird is the name Deborah Davis used in her 1979 biography of Washington Post owner Katharine Graham, for a network she said CIA covert-operations chief Frank Wisner built to recruit leading journalists. The Family Jewels don’t use that name, and Davis didn’t name her sources. Historian David Hadley concluded that Davis’s version — one systematic programme manipulating the whole US press — isn’t supported by the investigations, but stressed that this shouldn’t hide the CIA’s active role in influencing the press.26
So the activity people mean by “Operation Mockingbird” is documented: about 50 journalists with secret CIA relationships (Church Committee) and 400-plus who carried out assignments (Bernstein). What’s disputed is whether it was one named, centrally run operation. Either way, it’s bad for the agency.
In 1975, the committee’s chairman, Senator Frank Church, went on Meet the Press and warned that the government’s eavesdropping capability could be turned on its own citizens, leaving no American any privacy.4 He was talking about the NSA’s intercepts, and he didn’t know it yet, but the NSA had already monitored his own overseas calls.4
The press era needed people — recruited, managed and kept quiet one at a time. The platform era runs on a different lever: whoever writes the rules and tunes the recommendation systems decides what billions of people see. Here is what the record shows about who has been near that lever.
BLUE Government RED Platforms GOLD Disclosure DARK RED Australia
In July 2022, MintPress News reported that Meta had recruited dozens of people from the CIA, plus many more from the FBI and the Defense Department, mostly into trust, safety and content moderation. Among them: a former senior CIA analyst who became Meta’s senior policy manager for misinformation, a former FBI supervisory special agent who became a director of trust and safety, and a former USAID deputy policy director who became a misinformation policy manager.8 MintPress is openly adversarial to the US national-security establishment, and its evidence is LinkedIn profiles. But the profiles are specific and checkable, and the article itself stresses that it is describing a structural problem, not accusing individuals.8
Five months later, the first-hand evidence arrived. In Part 6 of the Twitter Files, Matt Taibbi published internal records showing more than 150 emails between the FBI and Twitter’s former trust-and-safety chief from January 2020 to November 2022, “soon to be weekly” meetings, and a stream of accounts flagged for “possible violative content” — including joke tweets from accounts with few followers.6 Part 8 showed Twitter had whitelisted accounts for US Central Command that later ran covert Arabic-language messaging.6
When states and users sued over the wider government pressure campaign, the Supreme Court ruled 6–3 in June 2024 that they lacked standing, because they couldn’t connect any government message to their own removal; three justices dissented.7 Our source for the Files’ contents is a sympathetic summary, which we flag as such.6 The fair reading: the contact was real and extensive; whether it amounted to coercion has never been decided on the merits.
On 21 November 2024 the government introduced the Online Safety Amendment (Social Media Minimum Age) Bill. The House passed it 102 votes to 13, and it cleared both houses on 29 November — eight days later.910 The law requires platforms to take “reasonable steps” to stop Australians under 16 holding accounts. There is no exception for parental consent.9
The campaign behind it is documented. News Corp Australia ran “Let Them Be Kids” across its mastheads after Meta declined to renew its news-payment deals in March 2024, and the Prime Minister thanked News Corp by name when announcing the law. The Senate inquiry took submissions for 24 hours.16 Tech Policy Press called it commercial logic operating in plain sight: the law’s targets are the same platforms that took the newspapers’ advertising.16
The original file argued that the agencies spent 60 years learning to shape adult beliefs through the press, then 20 years migrating that apparatus to social media — and that something unexpected happened. Teenagers weren’t being nudged towards compliance; they were becoming angrier and more distrustful of institutions, in directions nobody controlled. On this reading, the ban isn’t about protecting children; it’s about the algorithm no longer working on them.
We have no documents that show this motive. It is an inference from the pattern, and readers should weigh it as one. The documented motives are parental concern, a media campaign with a commercial interest, and a government that found the policy popular.
Platforms already set 13 as their minimum. Australia chose 16, which the Prime Minister described as giving teenagers “three more years” shaped by real life rather than algorithms.17 European Commission President Ursula von der Leyen said she was “inspired by Australia’s example”,17 and by March 2026 France, Germany, Spain, Greece and Denmark were considering national bans while the Commission’s expert panel met.16 On the other side, the Council of Europe’s human rights commissioner said bans shift responsibility from platforms to children, and 430 security and privacy scientists called for a moratorium on age-assurance technology until its harms are understood.16
Our March file said the ban would build the biggest identity database in Australian history. The Act points the other way. Platforms may not collect government ID or use Digital ID unless they also offer a reasonable alternative, and anything collected to check age must be destroyed after use — failing to destroy it is an interference with privacy under the Privacy Act.9 eSafety says platforms don’t have to put every account holder through an age check if they already have other accurate data.11
So how do platforms find the children? By watching everyone more closely. eSafety’s own list of “age signals” platforms may use includes facial age analysis of your photos and videos, voice analysis, the language you use, activity patterns that fit school timetables, and who you’re connected to, plus location signals such as IP address, GPS and device settings.11 Meta says it now uses AI to scan profiles — posts, comments, bios and captions — for clues such as birthday celebrations or mentions of school grades.12
That is the nudge machine turned into an inspection machine: the same behavioural profiling that powers the feed, now pointed at the question “how old is this person?” For how that kind of profiling builds a model of you, see Digital Twins.
Since March, the government has escalated rather than paused. It doubled the maximum penalty to A$109.2 million and gave eSafety power to compel documents and question current and former staff, and on 20 October 2026 the Communications Minister said eSafety was preparing Federal Court action against Facebook, Instagram, TikTok, Snapchat and YouTube.14 Meta says it has removed more than 750,000 Australian accounts and wants age checked once, at the phone’s operating system or app store.12 The government’s next step, a Digital Duty of Care bill, would cover services from social media to AI chatbots and let users switch off algorithmic feeds.15
From 1 December 2026, businesses can join the Australian Government Digital ID System and use myID to check who their customers are. Businesses that do will pay for it from 1 January 2027 at the latest; individuals and government services won’t.18 That raises an obvious question nobody has answered yet: will Facebook, Instagram and the other age-restricted platforms opt in, and use your government digital identity to decide whether you can have an account?
In fairness, a bank- or government-backed “yes/no” could be more private than letting a platform scan your face and read your posts — that is exactly the argument ConnectID makes: the platform learns only “over 16”, not your name or date of birth.21 The trade-off isn’t whether to verify; it’s who holds the link between your real identity and your online life. For where the myID system came from and how fast it’s growing, see The Day LifeLog Died.
Meta’s preferred fix — verify age once at the device or app-store level — would move the checkpoint from each app to the phone itself.12 That may well be more private than every app scanning your face. It would also mean the gate sits in two companies’ operating systems, in front of every app you install. Combined with a national Digital ID that opens to business on 1 December (see 1984 By Consent), these are the pieces of infrastructure to watch.
The Digital Duty of Care Bill went into Parliament on 15 October 2026. The minister dropped the power to define new harms by decree — new harms now need an Act of Parliament — and platforms get 12 months to comply once it passes.22 Its headline promise is “My Feed, My Way”: every user over 16 gets a prompt to choose between the algorithm’s recommendations and a feed of only the accounts they follow, and can change their mind as often as they like.23
But the bill itself doesn’t spell out how the choice works. It gives the minister power to make a separate “user empowerment tool rule” after the bill passes. A draft of that rule is out now; the department describes it as requiring social media services to offer logged-in users a following-only feed.24 So the details that matter — including which way the switch starts — are in a rule most people will never read.
The government is taking public submissions on the draft “My Feed, My Way” rule until 11.59pm AEDT, 5 November 2026. You can submit through the department’s Have Your Say page, or email the completed submission template to UERules@infrastructure.gov.au. You can choose to stay anonymous, or keep your name and submission private.24
Questions worth putting in a submission, in your own words:
Mockingbird-era tradecraft worked through the press. In They Own The Sky and Australia & The Ozone Hole we traced how a public message about the sky was shaped and repeated until it became common sense. The algorithm now does that work for the feed. And having let the feed run, governments are now deciding who may stand in front of it.
1. The CIA kept covert ties with about 50 US journalists until 1976 (Church Committee) and, by Bernstein’s count, 400+ journalists carried out assignments over 25 years.
2. The US military contracted software for fake online personas in 2011, aimed (it says) at foreign-language audiences.
3. Former intelligence and law-enforcement staff hold trust-and-safety roles at the major platforms, and the FBI was in constant contact with Twitter about specific accounts and posts.
4. Australia’s under-16 law passed in eight days after a media campaign by a company whose competitors it targets, with no parental-consent exception.
5. Enforcing it relies on platforms profiling users’ faces, voices, words and habits to estimate age — while most children already on the major platforms remain there.
That the ban built an ID database — it didn’t, and we were wrong to say so. That the CIA runs the platforms — the record shows hiring, contact and pressure, not control. That the Twitter Files proved government coercion — the courts never reached that question. That Europe is moving in lockstep — most countries are still considering. And that the ban’s real purpose is to recapture young minds — that is our hypothesis, and we have marked it as one. The parental concern behind the ban is genuine, and so are the platform harms it responds to. The question this file asks is who else benefits, and what gets built along the way.
This is information, not instruction — and not legal advice. If any of this bothers you, here is where your attention has the most effect.
◆ FIND US ◆ FOLLOW US ◆