Say It Out Loud: It’s the Platforms

The UK state still regulates controversial speech as though maximal influence were exercised principally by individuals in physical venues. Yet decisive power now lies with platforms that select, recommend, monetise and amplify speech on an industrial scale. Every night, virtual stadiums are filled through livestreams, recommendations and viral posts.

Their audiences are larger than almost any political rally, yet the machinery that fills them remains less visible and less scrutinised by authorities than the person on stage.

This article argues for regulating the amplifier rather than relying primarily on prosecutions and territorial exclusion. The focus will be on the UK, though with lessons for all democracies. As Belfast burns amid racially motivated violence, arson and attacks on police that Ofcom says appear partly to have been incited online, the weakness of the present settlement is deafeningly loud.^[1]

Britain prosecutes individual users and excludes controversial foreign speakers, yet meekly responds to the platforms distributing their messages with warnings, consultations and negotiated commitments. Even where freedom of speech is maximally protected, a right to algorithmic amplification need not be.

A regulatory model of yesteryear in a new information environment

Responding to another predictably inflammatory post from Elon Musk, the journalist Paul Mason recently asked on X: “More hate — what’s Ofcom doing?” The post demonstrates that, at least by some, scrutiny is shifting towards the failing machinery of platform oversight.

Britain’s institutions were built for an information environment of identifiable publishers, physical venues, finite audiences and national borders. A speaker entered the country, addressed a crowd, published through an editor and could be prosecuted when a clear legal threshold was crossed.

That model has not disappeared, but it is no longer dominant. A single stream can be watched live by tens of thousands, clipped into hundreds of fragments and recommended for days to audiences who never sought it out. Sensationalism is rewarded because outrage retains attention. Coordinated bot networks, hostile states and organised political actors can further strengthen an apparently organic surge.

This is not speculation about a hypothetical future. Parliament’s Intelligence and Security Committee has described Russia’s use of social media, disinformation and “hack-and-leak” operations, with the Government concluding that Russian actors sought to interfere in the 2019 general election through online amplification of stolen documents.^[2] Research on digital media and political polarisation remains contested in its details, but the broader evidence shows that online systems can alter exposure, hostility and the circulation of partisan material.^[3]

The defensible claim is not that algorithms created every political division. They themselves are morally agnostic. It is that they can identify, intensify and monetise divisions that malicious actors are then able to exploit given their bias to maximise engagement.

The state still reaches first for the individual: A blunt weapon for a task demanding precision

Britain has demonstrated that it can act quickly against individuals. During the 2024 public disorder, people were prosecuted and imprisoned for online posts intended to stir racial hatred or encourage attacks.^[4]Such action is justified where speech crosses established criminal thresholds.

The state also uses immigration powers against foreign speakers whose presence is judged “not conducive to the public good”. ^[5] Official guidance allows exclusion for conduct expressed through speeches, publications, websites or other media that foments terrorism, provokes serious criminality or fosters hatred.^[5]

This power has been applied across political tendencies. Pamela Geller and Robert Spencer were excluded before a planned English Defence League rally in 2013. Lauren Southern, Martin Sellner and Brittany Pettibone were refused entry in 2018. In 2026, Hasan Piker and Cenk Uygur were prevented from travelling to Britain for appearances including SXSW London and Oxford events.^[6][7][8]

These cases are not identical, and inclusion in the same list does not establish moral equivalence. They demonstrate a recurring method: identify the controversial personality, close the physical venue and present the exclusion as risk reduction.

But the modern venue does not close at the border—so far, at least.

Nevertheless, Piker and Uygur could still address the Oxford Union remotely. Piker could continue his long daily Twitch broadcasts, while material from him and countless comparable political streamers remained available through Twitch, YouTube, X and clips circulated across other services.^[8] The Government prevented bodies from entering Britain while leaving the principal channels through which their speech reaches British audiences intact.

The Google Trends data provide a useful illustration. Search interest in Hasan Piker in Britain rose sharply around the 2026 exclusion, while his earlier physical visit produced no comparable surge. This does not prove that the ban persuaded anyone, nor that every search became a follower. It does show that exclusion can become an amplification event: an obscure or limited appearance is transformed into a national controversy.

Imagine that the state prevents a speaker from addressing 500 people but the resulting controversy brings the same speaker before 20,000 online viewers. The smaller venue has been closed while the larger one has been advertised.

The platforms are the venues

Platforms describe themselves as hosts, but a major platform is simultaneously the venue owner, promoter, advertising agency, ticket distributor and loudspeaker. The greater the number of views, the greater the revenue, regardless of the influence and consequences of the content.

They do not merely store speech. They select what appears in feeds, suggest what plays next, identify audiences likely to engage and sell advertising against the result. Engagement trumps all else.

The creator produces the speech but the platform dictates its scale and spread.

This creates an obvious strategic inconsistency. How can a person be considered sufficiently dangerous, hateful or inflammatory to justify exclusion from Britain, yet their videos remain available on platforms whose search and recommendation systems can deliver them to British users? How can their daily streams continue forming new audiences without any corresponding public intervention from the regulator Ofcom?

There is no automatic legal contradiction. As of the writing of this article, the Home Office’s public-good test is broader than the criminal thresholds underlying the Online Safety Act. A person can be excluded without every statement they make being illegal, and Ofcom regulates platforms’ systems rather than deciding which named individuals must be removed.

Nevertheless, the strategic contradiction remains. If the state believes an individual’s physical appearance creates a serious risk, that should at least trigger scrutiny of the much larger digital venues carrying the same material. If the law is intended to reduce harm from incitement, then why ignore the far more toxic incitement that is found online in substantially larger measure?

Yet Ofcom’s published response to major platforms has often been generalised pressure rather than visible, case-linked accountability. Even after gathering evidence that terrorist content and illegal hate speech persisted on some of the largest services, Ofcom accepted commitments from X concerning review times, reporting systems and accounts linked to proscribed organisations rather than announcing immediate sanctions.^[9]

Ofcom was right to state during the Belfast unrest that it does not order the removal of particular posts or accounts. But this cannot become an excuse for regulatory abstraction. Systems are experienced through particular cases. If repeated videos, streams or posts present an obvious test of a platform’s risk assessments, moderation and recommender systems, Ofcom should be able to ask what the platform did, when it did it and whether it continued to recommend or monetise the material while reviewing it.

The Government should not be able to declare a speaker too dangerous for a hall in Britain while treating the virtual stadiums hosting that speaker as someone else’s problem.

Regulate amplification, not opinion

The choice is not between doing nothing and deleting every controversial speaker. A proportionate system could, in most cases, preserve the ability to publish while restricting the platform’s active role in creating new audiences.

An account placed under an amplification restriction might remain available to existing followers, direct links and exact-name searches. It could still speak. But it would be removed from trending lists, autoplay, broad search recommendations and recommendations to non-followers. Monetisation could be suspended, and clips or mirror accounts could be prevented from evading the restriction.

This distinction matters. Freedom of speech protects the ability to express and receive ideas. It does not obviously include a right to have a company’s prediction systems recruit an audience, place the content before people who did not request it and pay the speaker for the resulting engagement.

Such restrictions would still affect practical freedom of expression and therefore require safeguards: published thresholds, written reasons, independent appeals, time limits, regular review and full disclosure of government requests. The rule must address conduct and legally recognised risk, not political viewpoint.

Fair enforcement does not mean engineering an equal number of sanctions against left and right. It means applying the same threshold to Islamist propaganda, antisemitism, anti-Muslim agitation, racial incitement and encouragement of political violence.

There is also room for an intermediate category between legal content and proven criminality. Ofcom need not declare a political opinion illegal before requiring a platform to examine whether repeated recommendation of the material creates a foreseeable systemic risk. Demotion is not imprisonment. Demonetisation is not censorship. Removal from autoplay is not a prohibition on speaking.

These measures introduce friction where platforms currently manufacture momentum.

In the UK, the fundamental legal levers already exist

The Online Safety Act 2023 already requires regulated services to assess illegal-content risks, remove illegal material swiftly once aware of it and reduce the likelihood that users encounter it.^[10] Ofcom, the oversight body,  may impose substantial fines and, in serious cases of continued non-compliance, seek court-ordered service or access restrictions.^[11]

The weakness lies less in the absence of power than in enforcement priorities and regulatory confidence. Ofcom has used formal sanctions against pornography providers, 4chan and a suicide forum. It has been more hesitant in the politically explosive territory of illegal hate, terrorist rhetoric and the recommender systems of the largest platforms.^[9][12]

Letters and voluntary commitments may begin a process, but they are not an adequate endpoint where the regulator itself says illegal material is persisting.

The European Union’s Digital Services Act offers a complementary model. Its largest platforms must assess and mitigate systemic risks, including risks arising from recommender systems, and submit to stronger transparency and audit duties. ^[13]The European Commission’s €120 million fine against X in 2025 concerned transparency failures rather than hate speech, but it demonstrated that the EU is prepared to impose a major platform-level penalty. ^[14]

Britain should combine the strongest elements of both regimes: the EU’s emphasis on systemic risk, auditing and recommender systems with the Online Safety Act’s severe business-disruption powers.

The current weak link is Ofcom. Ofcom should be able to require platform-specific amplification controls, independent algorithmic audits and rapid crisis protocols. Enforcement could escalate from formal warnings and mandatory audits to demonetisation and removal from recommendations, followed by temporary suspension and, only in persistent cases, service restrictions.

This would shift regulation upstream. Rather than waiting for thousands of individuals to spread illegal or inflammatory material and then prosecuting a handful of them, the regulator would address the systems that selected the material for mass distribution.

The future of Ofcom, a need for greater clout

Ofcom’s present weakness should not be mistaken for proof that platform regulation cannot work. The illegal-content duties only became enforceable in March 2025, and the regulator has already shown in cases involving 4chan and a suicide forum that it can move beyond guidance and warning letters to investigations, financial penalties and remedial orders. ^[12][15] Its stated priorities for 2026/27 now include countering terrorism and illegal hate, bringing its lawyers, technologists, researchers and enforcement staff together around those areas. ^[16]There is, therefore, reason for cautious hope: the machinery exists, the first precedents have been established and the regulator has begun to identify the correct targets.

The problem is that the machinery still moves at the speed of conventional regulation while online amplification moves at the speed of a viral post. Ofcom’s enforcement process requires evidence-gathering, provisional findings, representations from the company and a final decision. ^[17]Those safeguards are essential where restrictions may affect political speech, but they can also mean that millions of impressions have already occurred before a case reaches any meaningful conclusion. A regulatory system designed only to punish a platform months later may document the harm without preventing it.

Ofcom should therefore be strengthened in practical, limited ways.

• First, it should have a dedicated rapid-response unit for periods of civil disorder, terrorist attacks and elections. Where there is credible evidence of illegal hate, incitement or coordinated foreign manipulation, that unit should be able to demand the preservation of platform data, an immediate account of how recommendation systems have distributed the material and a temporary halt to monetisation or automated recommendation while the evidence is assessed. 
• Second, Ofcom should force platform-specific transparency rather than accepting general claims about moderation. Major services should publish comparable data showing how quickly reported illegal hate and terrorist material is assessed and similar questionable media. Voluntary commitments should be converted into measurable benchmarks with fixed deadlines. In persistent cases, the business-disruption measures already available under the Online Safety Act should be utilized. Platforms must not be able to brush off the rules with effective payoffs.
• Third, Ofcom should be required to examine the connection between territorial exclusion and digital reach. A Home Office decision that a speaker’s presence is not conducive to the public good should not create an automatic online ban. However, such a decision should trigger a concrete swift review of whether the relevant platforms have assessed the individual’s content, whether potentially illegal material is being monetised or recommended, and whether the same standards are applied to comparable figures from other political tendencies.

There is also a statutory limit that stronger enforcement alone cannot solve. Parliament removed the proposed duties concerning content that is legal but harmful to adults. ^[18]Ofcom can act against illegal material and against platforms whose systems fail to manage the risks identified in the Act. It does not possess a general mandate to restrict lawful yet inflammatory political content merely because its amplification creates a foreseeable social risk.
Parliament should consider a narrow power directed at amplification rather than publication. Reviewable restrictions on recommendation and monetisation where a platform’s systems are demonstrably intensifying a serious and immediate risk. Reach should be a critical part of the calculus.

The reason for hope is that the Online Safety Act already recognises the platform as a responsible actor rather than a passive noticeboard. The next step is to make that responsibility timely, measurable and enforceable through a strengthened potent Ofcom. The UK is almost there.

A footnote on an example of stricter platform regulation

Moldova offers a limited but useful example of earlier, platform-directed intervention. Before its 2024 presidential election and constitutional referendum, Moldovan authorities and European Commission officials conducted an election-readiness exercise with platforms, civil-society groups, fact-checkers and independent media.
Several platforms agreed to a temporary escalation mechanism for election-related threats. ^[19]Meta subsequently removed a coordinated network of fictitious Russian-language news brands targeting Moldovan audiences, stating that it had acted before the network could establish authentic audiences on Facebook and Instagram.^[20] The intervention did not depend on proving that every individual post was false or criminal. It targeted coordinated inauthentic behaviour: the artificial manufacture of popularity and legitimacy before platform algorithms could convert it into genuine reach.

The comparison should not be overstated. Moldova faced an acute foreign-interference campaign surrounding a specific election, whereas Ofcom regulates a much larger and permanent information environment. Meta’s account is also a company’s description of its own enforcement, and international observers still found that disinformation and foreign interference damaged the integrity of the electoral environment. ^[21]
Moldova therefore does not demonstrate that stricter regulation can eliminate hostile information campaigns. Perfection is unrealistic. It supports a narrower claim. Authorities, independent monitors and platforms can establish rapid channels before a crisis, identify manipulation through behaviour rather than viewpoint, and interrupt artificial audience formation before it becomes self-sustaining.

 In this era of Russian troll factories, Britain could heed the Moldovan lesson with Ofcom coordinating trusted reporting, urgent data requests and temporary restrictions on automated amplification.

A need for urgency

As violence unfolds in Belfast, Ofcom has warned platforms that ordinary moderation may be inadequate during a crisis and that viral illegality can translate into real-world harm.[1] ^[1]That recognition should lead to a substantial bolstering of measures, with rapid implementation.

Britain cannot plausibly treat a lecture hall as the decisive venue while Twitch, YouTube and X fill virtual stadiums every night. It cannot rely on border exclusions that may amplify the people they target, then regard the systems delivering those same people to vastly larger audiences as a secondary concern.
No more light touches.

The state should be slow to silence individuals curtailing criticisms of suppressing free speech. Instead the lens should be on the machinery that manufactures their reach and the devastating consequences that could lead to the failure of UK democracy.