In an age in which a rumour can cross a continent before a newspaper has finished checking its first source, the boundary between legitimate public debate and organised manipulation has become increasingly difficult to police. From Westminster to Brussels and Washington, democratic governments face a problem that did not exist at anything like this scale a generation ago: how do you confront organised disinformation without giving governments, regulators or technology companies excessive power over what citizens are permitted to say, read and debate?
It is a genuine dilemma. Disinformation is not simply a matter of somebody getting a fact wrong on the internet. Coordinated campaigns can exploit social divisions, manufacture outrage, impersonate trusted sources and manipulate the systems that determine what millions of people see. Online activity can also have consequences offline, from undermining confidence in public institutions to intensifying tensions during elections or crises.
That is why governments have acted. In Britain, the Online Safety Act places duties on regulated online services to assess and manage risks associated with illegal content, while requiring transparency and protections for freedom of expression. Ofcom's role is principally to ensure that platforms operate appropriate systems; it does not ordinarily instruct companies to remove individual posts or accounts. In the European Union, the Digital Services Act has adopted a broader systemic approach. Very large platforms and search engines must assess risks arising from their services, including risks to electoral processes, fundamental rights and civic discourse, and take measures to mitigate those risks.
The principle behind these laws is understandable. The difficult question is what happens when responsibility for managing those risks is ultimately translated into decisions made by private companies and, increasingly, automated systems.
THE DANGER OF DELEGATED MODERATION
The modern internet is governed partly through algorithms. Platforms decide what is recommended, what is demoted, what is labelled, what is investigated and, in some circumstances, what is removed. Regulation can therefore influence not only whether illegal material is taken down, but also how companies design the systems through which lawful speech reaches an audience.
That creates an uncomfortable possibility. A platform facing regulatory obligations and substantial penalties may have an incentive to minimise risk. Where the boundary between harmful material and lawful but controversial speech is difficult to establish, an overly cautious system can become attractive from a corporate perspective.
The result need not look like traditional censorship. There may be no government official ordering a newspaper to stop publishing an article. Instead, an automated system may reduce its visibility, remove an account, restrict advertising or classify material as risky.
That distinction matters. A democratic society should be extremely cautious about creating a system in which lawful political argument is effectively filtered by commercial algorithms simply because those algorithms are designed to avoid regulatory risk.
The European model itself recognises this tension. The Digital Services Act contains safeguards relating to freedom of expression and media freedom alongside its requirements for platforms to address systemic risks. The challenge, therefore, is not simply whether governments should regulate technology companies. It is how that regulation is designed, how it is enforced and what safeguards exist when reasonable people disagree about the meaning of contested information.
NOT EVERY FALSEHOOD IS A DISINFORMATION CAMPAIGN
There is another distinction that deserves much greater attention. Misinformation generally refers to false or inaccurate information shared without an established intention to deceive. Disinformation normally refers to deliberately deceptive information, particularly when it is organised or coordinated.
The distinction is not always easy to establish in practice, but it is important. A citizen can be wrong without being malicious. Someone can believe an inaccurate claim about economics, medicine, history, immigration or foreign policy and publish it online. They may be misinformed, careless or simply mistaken. In an open society, the answer to that kind of error is not automatically censorship.
Democracy has always depended on the ability of citizens to be wrong. Many beliefs once regarded as eccentric or unacceptable have eventually entered legitimate public debate. That does not make every unconventional claim true. It means that the state should be very careful before deciding that an opinion is too misguided to be heard.
The situation is different when there is evidence of a coordinated operation designed to deceive at scale: networks of automated or inauthentic accounts, fabricated identities, coordinated manipulation, covert political advertising or deliberate exploitation of recommendation systems. There, the focus should be on conduct, coordination and infrastructure rather than on the mere existence of an unpopular opinion.
That is a much safer principle for a democracy.
FOLLOW THE MACHINERY, NOT JUST THE MESSAGE
The most useful regulatory question may therefore be changing. Instead of asking simply, “Should this piece of content be removed?”, policymakers should increasingly ask: Who created it? Who is funding it? How is it being distributed? How quickly is it being amplified? Are automated or coordinated networks involved? And is the platform's own architecture contributing to its reach?
That approach moves attention from individual speech towards the systems capable of manipulating millions of people.
There is already movement in this direction. The EU's Digital Services Act requires very large platforms and search engines to assess systemic risks associated with their design and operation, including coordinated disinformation campaigns and risks to electoral processes. The EU has also incorporated a strengthened Code of Conduct on Disinformation into the DSA framework, with an emphasis on transparency and freedom of expression. Britain is also moving towards greater scrutiny of platform systems rather than simply individual pieces of content. Ofcom's online-safety regime places considerable emphasis on risk assessments, systems and processes, record-keeping and transparency.
That is potentially the more sustainable direction.
THE CASE FOR ALGORITHMIC TRANSPARENCY
If algorithms increasingly influence what citizens see, democratic societies have a legitimate interest in understanding how those systems operate. That does not necessarily mean publishing commercially sensitive source code. It can mean independent auditing, meaningful transparency about recommender systems, access for qualified researchers, explanations of major content-moderation decisions and proper mechanisms for users to appeal decisions affecting their speech.
The objective should be accountability without turning regulators into editors.
That distinction is crucial. A regulator should be able to ask whether a platform has adequate safeguards against coordinated manipulation. It should be able to examine whether its systems systematically amplify demonstrably harmful activity. It should be able to demand transparency where the law permits.
But the further the regulator moves towards determining which lawful political arguments citizens may encounter, the greater the democratic risk becomes.
FREE SPEECH IS NOT A DEFENCE OF FALSEHOODS
There is an important point on the other side, too. Defending freedom of expression does not mean pretending that disinformation is harmless.
False information can cause real damage. Coordinated manipulation can distort public debate. Fabricated material can damage individuals and institutions. During a crisis, rapidly spreading illegal content can contribute to real-world harm. Nor should “free speech” become a convenient shield for foreign influence operations, fraudulent networks or deliberately deceptive campaigns conducted at industrial scale.
The answer, however, should be proportionate. A mature democracy must be capable of distinguishing between a citizen expressing a mistaken opinion and an organised operation attempting to manipulate an entire information environment. Those are not the same problem, and they should not automatically receive the same regulatory response.
FROM CONTENT POLICING TO DEMOCRATIC RESILIENCE
The long-term answer is unlikely to be found in deletion alone. It requires a wider strategy.
First, improve algorithmic accountability. Independent researchers and appropriately authorised regulators should have meaningful ways to examine how large platforms recommend, amplify and moderate information.
Second, protect legitimate journalism and political debate. Platforms need effective appeals and safeguards against automated systems wrongly suppressing lawful material.
Third, attack coordinated manipulation at its source. Fake accounts, bot networks, deceptive advertising, coordinated inauthentic behaviour and undisclosed influence operations deserve greater scrutiny than an ordinary citizen who happens to repeat something untrue.
Fourth, invest in media literacy. A population trained to question sources, recognise manipulated images, examine evidence and distinguish reporting from propaganda is harder to deceive than one that simply waits for a government agency or technology company to decide what is true.
And finally, keep independent journalism alive. The best antidote to a polluted information environment is not necessarily a more powerful censor. It is a stronger information ecosystem: professional journalism, transparent evidence, competing viewpoints, credible institutions and citizens capable of thinking critically about all of them.
THE DEMOCRATIC TEST
The great danger of the digital age is not only that people may believe false things. It is that governments, companies and citizens may gradually lose confidence in the ability of an open society to deal with falsehood without controlling speech.
That would be a serious mistake.
Democracy cannot promise its citizens an internet free of lies. It cannot guarantee that every viral claim will be true. Nor should it attempt to create a Ministry of Truth for the digital age.
What it can do is build institutions capable of exposing deception, scrutinising powerful platforms, disrupting organised manipulation and correcting falsehoods while leaving sufficient space for citizens to argue, disagree and, occasionally, get things spectacularly wrong.
The distinction is fundamental.
A democracy should regulate the machinery of manipulation more carefully than it regulates the existence of opinion.
The algorithm should not become the new censor. Nor should the state.
The objective should be something more difficult and more valuable: a digital public sphere in which deception can be challenged, powerful platforms can be held accountable and citizens remain free to think, speak and disagree.
Because democratic resilience is not achieved when everyone is forced to hear the same thing. It is achieved when citizens are free to hear competing arguments — and have the knowledge, evidence and institutions necessary to decide what deserves to be believed.