Democracies across the world are assembling the machinery of media control that we once associated with totalitarian states. They all say it’s for our own good.
Misinformation and disinformation are real problems, and they are getting worse.
Generative AI can fabricate a plausible video of anyone saying anything. Algorithms reward outrage over accuracy. Deepfakes circulate faster than corrections. Hostile states run industrial-scale influence operations.
The British communication regulator, Ofcom, found that only 45% of UK adults feel confident judging whether a source of information is truthful. The BBC found that AI assistants garble the news they summarise about half the time.
This problem isn’t new, of course. Both governments and religious authorities responded to the invention of the printing press with great alarm, recognising its potential to spread “misinformation”, as they described information that undermined their authority. This spawned sustained efforts to control the output of printers by mandatory licensing regimes and requiring pre-publication approval of printed texts.
Still, if you were designing an environment in which lies outran truth, you could hardly improve on the modern internet.
Controlling instincts
Today’s governments respond with the same controlling instincts that motivated the kings and popes of yore. They justify their need to control information with claims that powerful new technologies produce harms (“fear the terrorists!”, “think of the children!”, “how will people know what the Truth™ is?”) that private actors will not voluntarily prevent, so the state must “do something”.
The trouble is that the state is itself not a trustworthy actor.
Survey the democratic world in mid-2026, and a pattern emerges that ought to alarm anyone who values a free society: governments are converging, from different directions and with different excuses, on ownership or control of the information available to their citizens.
Arbiter of trust
Start with the United Kingdom, where the Department for Culture, Media and Sport has published a green paper, Watch this space, on the future of British media.
It is a masterclass in the language of benevolence. The government proposes to “foster a healthy information environment” by exploring legislation to force social media platforms to make “trustworthy news” prominent and easily discoverable – with public service broadcasters first in the queue – and to impose new “media literacy” duties on broadcasters.
The word “trustworthy” does an extraordinary amount of work in this document, and it is doing that work dishonestly. Trustworthiness, in ordinary usage, is a judgement that consumers of media render upon producers of media. It is earned, granted and revoked by readers and viewers, one at a time.
In the green paper, it means something else entirely: it means what the government deems trustworthy. The paper admits as much, conceding that criteria for a “trustworthy news provider” have not yet been determined and will be developed by the state in consultation with “stakeholders”.
Note the sleight of hand. Trust in UK news media has fallen to 35%, the paper laments. But trust in the government is lower still. A government that the public does not trust proposes to decide, on the public’s behalf, which news sources the public ought to trust, and to compel private platforms to push those sources into everyone’s feeds.
“Further responsibilities”
The paper even floats making the benefits of prominence conditional on “further responsibilities” – that is, on behaving in ways of which the state approves. It is less a trust framework than a control regime for both professional and amateur media, with algorithmic prominence as the carrot and algorithmic obscurity as the stick.
The same government, in the same month, banned social media for everyone under 16, with curfews and scroll-limits for older teenagers under consideration.
I have written before about why kicking kids off social media is both futile and harmful. The relevant point here is that a ban on minors requires age verification of everyone – which means identity verification of everyone, which means the end of anonymous speech for adults.
Reading your email
If Britain wants to curate what you read, the European Union wants to read what you write.
On 9 July, the European Parliament allowed “Chat Control 1.0” – the suspicionless mass scanning of private communications – back into force until 2028. It did so despite the fact that a majority of voting members of the European Parliament opposed it, 314 to 276, simply because the motion to reject it fell short of the absolute-majority threshold of 361 votes. A measure the elected chamber twice rejected in March now stands, because not enough parliamentarians could be bothered to turn up for work.
The ostensible reason is child protection, a cause so unimpeachable that it has become the universal solvent of civil liberties.
But the numbers, most of them from the EU Commission’s own reports, dissolve the pretext. Mass scanning of private chats produced barely a third of abuse reports in 2024.
The German federal police found that nearly half of incoming alerts were not criminally relevant at all, and 40% of resulting investigations targeted minors themselves – teenagers sexting each other, criminalised by the machinery supposedly built to protect them.
The Commission concedes there is no evidence that suspicionless scanning has increased convictions or rescued children.
Encrypted channels exempt
End-to-end encrypted channels are exempt from the EU’s snooping law, not because they wouldn’t like to read them, but because they technically cannot do so. Meanwhile, every actual predator with a functioning brainstem long ago switched to encrypted platforms.
What remains under surveillance is the correspondence of the law-abiding – your Outlook or Gmail, your social media direct messages, your in-game chats – scanned without warrant or suspicion, in flat contradiction of the presumption of innocence and the centuries-old principle that the state does not open your mail unless a judge says it may.
As abuse survivor and privacy advocate Alexander Hanff put it: “We survivors need privacy, because without it we lose our voice.”
Mass surveillance destroys the very safe spaces in which abusers can speak up, support each other, and seek justice. Surveillance that criminals can trivially evade is not crime-fighting. It is the dismantling of privacy as such.
Liability as censorship
Lest we feel smug at this distance, South Africa is assembling its own apparatus, piece by piece.
The Films and Publications Amendment Act – administered by the Film and Publication Board, the direct institutional descendent of apartheid’s Publications Control Board – already requires online distributors to register with the state and submit content for classification, and criminalises loosely defined “harmful” content that causes “emotional, psychological or moral distress”. (I guess horror movies or documentaries about rape are a no-no, then.)
Now the Competition Commission’s Media and Digital Platforms Market Inquiry recommends amending the Electronic Communications and Transactions Act to make platforms legally liable for harmful content that includes mis- and disinformation, and to oblige platforms to proactively remove harmful content, under the eye of a new Information Ombud.
Age restrictions requiring robust identity verification are reportedly on the table too, following in the footsteps of nanny states abroad.
Consider the incentives this creates. A platform facing liability for content it fails to remove, but no penalty for content it wrongly removes, will always err on the side of deletion.
It will not hire South African lawyers to adjudicate whether a given post about, say, farm murders, or BEE, or illegal immigration, constitutes “misinformation”. It will build a classifier, set the threshold conservatively, and let the machine mow down lawful speech by the hectare.
Pre-emptive, precautionary censorship is not a bug of intermediary liability; it is the design. And who decides what counts as misinformation? Not a court, applying law to fact after the fact. A compliance department in a low-wage foreign jurisdiction, reacting to corporate fear of liability.
The state buys in
The United States, whose First Amendment forecloses the European route, has found another: direct ownership.
OpenAI has proposed handing Washington a 5% stake in the company – some $42.6 billion – as part of a scheme under which the government would hold 5% of every leading American AI developer through a sovereign wealth fund.
While the US president spent the Fourth of July warning that communism was a cancer to be cut out fast, his administration accumulated stakes in nearly two dozen companies, took 10% of Intel, holds a golden share in U.S. Steel, skims Nvidia’s China revenues, and promotes price-controlled “Freedom Fuel” petrol stations.
Senator Rand Paul asked the obvious question: if socialism is government ownership of the means of production, what exactly do we call this? Senator Bernie Sanders looked on with approval, and said the 5% stakes should be increased to 50%.
When socialists are applauding you, you’re probably doing something socialist.
The proximate cause of OpenAI’s tribute is instructive. Washington had just slapped export controls on Anthropic’s frontier models so sweeping that the company pulled them from public access entirely. After Anthropic agreed to significantly more restrictive guardrails, the export controls were lifted.
The message to the AI industry was unmistakable, however: the state can and will switch off your best product. The industry’s response – offering the state equity, or in Nvidia’s case, a cut of the revenue – is a pure protection racket.
So the US government has been acquiring a proprietary interest in the most powerful information tools citizens have ever had, alongside the regulatory power to decide who may use them and how capable they may be.
A state that is simultaneously shareholder, regulator and customer of the means of producing information is not an impartial referee. It is totalitarian.
Fit arbiter
Four jurisdictions, four mechanisms – curation, surveillance, liability, ownership – and one presumption: that government is a fit arbiter of what citizens may know, do, say and read.
The classical liberal tradition rejects that presumption root and branch, and not because misinformation is harmless.
John Stuart Mill’s argument in On Liberty did not depend on false opinions being rare or benign. It depended on three observations that remain true. The silenced opinion may be right, and the censor, being fallible, cannot know it is not. It may be partly right, and truth emerges only from collision with error. And even where it is wholly wrong, a truth defended by authority rather than argument is dogma, held without understanding, and abandoned for a plausible alternative at the drop of a tinfoil hat.
Hayek, in The Use of Knowledge in Society, described the knowledge problem: the information necessary to make rational decisions is dispersed, tacit, and local, making it impossible for any central authority to access, aggregate, use, or evaluate it effectively.
He applied it to central economic planning, but it applies equally to the question of determining the validity of information.
And not only is the state theoretically unable of deciding what is and is not misinformation, but it also has its own vested interests which make it an untrustworthy arbiter of truth.
Governments are themselves the largest and most motivated producers of misinformation on earth, with every incentive to define the term to exclude their own output and include that of their critics.
Editor-in-chief of reality
Even if we choose to trust today’s government, every censorship power created for ostensibly benevolent ends will be inherited by tomorrow’s government, which may not be benevolent.
South Africans, of all people, should not need reminding of this. We have lived under a state that decided which publications were trustworthy, which communications could be intercepted, and which information citizens could be permitted to see. It called that system moral protection too.
Comprehensive control over the information environment is not an incidental feature of authoritarian regimes; it is their defining one. Free societies are distinguished precisely by their refusal to install an editor-in-chief of reality.
The remedy for bad speech remains more speech, better speech, and citizens treated as adults – free to weigh, to doubt, to verify, and yes, to be wrong. The moment a government claims the authority to protect you from error, it has claimed the authority to define truth.
No government that has ever held that power has deserved it, and none ever will.
[Image: A cartoon by Holmet in the Feb-May 1916 issue of Motion Picture Magazine, depicting public sentiment threatening to swamp local, state and national censors. Public domain image.]
The views of the writer are not necessarily the views of the Daily Friend or the IRR.
If you like what you have just read, support the Daily Friend