# Social Media Regulation Is Long Overdue—And Must Go Further > The Online Safety Act 2023 is a start, but it doesn't address the fundamental business model driving harm: algorithmic amplification of outrage for profit. Real regulation means treating platforms as publishers, not neutral conduits. *Section: Opinion — By Naomi Clarke (Opinion Editor) — Published August 19, 2025 — 7 min read* Canonical URL: https://dailyjunction.org/opinion/social-media-regulation-overdue Tags: social media, technology, regulation, online safety, media, opinion ## Key takeaways - The Online Safety Act 2023 places new duties on platforms but stops short of treating them as publishers - Internal Meta research leaked in 2021 showed Instagram's algorithm knowingly harms teenage mental health - Algorithmic amplification prioritises engagement over accuracy, systematically promoting divisive content - Section 230-style immunity in the US has allowed platforms to avoid liability for content they actively promote - Ofcom now has enforcement powers but lacks the resources and mandate to fundamentally reshape platform incentives Social media platforms have spent the better part of two decades insisting they are neutral technology companies, not publishers, and therefore should not be held responsible for the content they host. That argument was always dubious. In 2025, after years of evidence showing platforms actively shape what billions of people see through algorithmic curation designed to maximise engagement at any cost, it is simply false. The Online Safety Act 2023 is a start, but it does not go nearly far enough. Real regulation means treating platforms as what they are: publishers with editorial control, and holding them to the same standards of liability we apply to every other media organisation. ## What the Online Safety Act actually does The Online Safety Act, which came into force in 2024, places new duties on platforms to remove illegal content, protect children, and assess risks from harmful but legal material. Ofcom, the communications regulator, now has powers to fine platforms up to 10% of global revenue for serious breaches. On paper, this is significant. In practice, it is a framework built on the assumption that platforms are fundamentally well-intentioned actors who just need clearer rules. It is not. The evidence—from internal research leaked from Meta, from whistleblower testimony, from academic studies of algorithmic behaviour—shows that the core business model of major platforms is to maximise engagement by promoting content that provokes strong emotional reactions, regardless of whether that content is true, harmful, or socially destructive. The Act does not touch that business model. It requires platforms to moderate content, but it does not require them to stop algorithmically amplifying the most divisive, misleading, and harmful material because that material drives engagement and therefore profit. It is like asking a tobacco company to put health warnings on cigarettes while letting them continue to market to children. It addresses symptoms, not causes. ## The evidence: platforms know what they are doing We do not have to speculate about whether platforms understand the harm their algorithms cause. We have the receipts. Internal Meta research, leaked in 2021 as part of the Facebook Files, showed the company's own studies found that Instagram's algorithm made body image issues worse for one in three teenage girls, and that teens blamed Instagram for increases in anxiety and depression. Meta knew this and did not change the algorithm, because doing so would reduce engagement. Similar patterns have been documented across platforms. YouTube's recommendation algorithm has been shown to systematically push users toward more extreme content—not because of a conspiracy, but because extreme content generates more watch time. X (formerly Twitter) amplifies outrage because outrage drives quote-tweets and replies, which the algorithm interprets as engagement. TikTok's "For You" feed is optimised purely for keeping users scrolling, with no regard for the accuracy or impact of what it shows them. > "These are not neutral platforms. They are engagement-maximisation machines that have learned, through billions of data points, that the fastest way to keep people scrolling is to show them content that makes them angry, anxious, or outraged. That is not a bug. It is the business model." — A conclusion now supported by extensive academic research into algorithmic amplification. ## The publisher question: if it curates, it is liable The legal fiction that platforms are not publishers rests on the idea that they are passive conduits, like a telephone network. That made sense in the early internet, when platforms simply hosted whatever users uploaded in chronological order. It makes no sense now. Modern platforms do not show you everything your friends post in the order they posted it. They use algorithms to decide what you see, in what order, and how prominently. They A/B test different versions of feeds to see which keeps you engaged longer. They promote content from accounts you do not follow because their data suggests you will interact with it. This is editorial curation. It is what publishers do. If a newspaper publishes a defamatory article, it is liable. If a broadcaster airs harmful content, it faces regulatory consequences. The principle is simple: if you exercise editorial control over what your audience sees, you bear responsibility for what you choose to show them. Platforms exercise that control at a scale no traditional publisher ever could, and they have been allowed to do so with almost no accountability. The Online Safety Act does not change this. It still treats platforms as a special category, subject to duties to moderate but not to the same liability as publishers. That needs to end. If your algorithm decides to promote content, you are responsible for that decision. ## The free speech red herring The inevitable objection is that this threatens free speech. It does not. Free speech means the government cannot criminalise expression. It does not mean private companies must amplify whatever anyone says, or that they are immune from liability for what they choose to promote. Newspapers have editorial freedom and are also subject to libel law, privacy law, and regulatory standards. They manage both. Platforms can too. The difference is that platforms have been allowed to claim the benefits of being publishers—curating content, selling targeted advertising based on that curation—without accepting the responsibilities. Requiring platforms to be liable for content they algorithmically promote does not mean they have to remove everything controversial. It means they have to make editorial judgements, just like every other publisher, about what is worth amplifying and what is not. If they choose to amplify defamatory, harmful, or illegal content, they should face the same consequences as anyone else who makes that choice. ## What real regulation looks like Effective regulation of social media requires three things the Online Safety Act does not deliver. **First, algorithmic transparency.** Platforms should be required to disclose how their algorithms work, what they optimise for, and what trade-offs they make between engagement and other values like accuracy or user wellbeing. This does not mean publishing source code. It means independent audits and public reporting on algorithmic impacts. **Second, liability for amplification.** If a platform's algorithm promotes content, the platform should be liable for that content in the same way a publisher is liable for what it prints. Hosting user content is one thing; actively choosing to amplify it to millions of people is another. **Third, a duty to offer chronological, non-algorithmic feeds.** Users should have the right to see content from accounts they follow in the order it was posted, without algorithmic interference. Platforms can offer algorithmic feeds as an option, but they should not be able to force users into engagement-maximising feeds with no alternative. None of this is technically difficult. It is politically difficult, because platforms are powerful, wealthy, and have spent years lobbying to preserve their immunity from liability. But the evidence of harm is now overwhelming, and the current regulatory framework is not fit for purpose. ## The mental health crisis we cannot ignore The stakes are not abstract. There is now a substantial body of evidence linking heavy social media use, particularly among teenagers, to increased rates of anxiety, depression, and self-harm. The mechanisms are well understood: social comparison, cyberbullying, sleep disruption, and algorithmically curated feeds that prioritise content designed to provoke anxiety and insecurity because that content drives engagement. This is not to say social media causes all mental health problems, or that everyone who uses it is harmed. But the population-level effects are real and measurable. Rates of anxiety and depression among UK teenagers have risen sharply since the mid-2010s, the exact period when smartphone ownership and social media use became near-universal. The correlation is not proof of causation, but combined with the internal research from platforms themselves, the case is strong. We would not allow a food company to sell products it knew caused harm and then claim immunity because it is just a platform connecting buyers and sellers. We should not allow social media companies to do the equivalent. ## The bottom line The Online Safety Act is a step forward, but it is not enough. It treats platforms as a special case requiring bespoke regulation, when the real solution is simpler: treat them as publishers. If they curate content through algorithms, they are making editorial decisions and should be held to editorial standards. If they amplify harmful content for profit, they should be liable for the harm. And if they cannot run a profitable business without algorithmically promoting misinformation, outrage, and content that harms children, then their business model is the problem and needs to change. The evidence is in. The time for half-measures is over. ## Frequently asked questions ### Doesn't regulating social media threaten free speech? No. Free speech means the government cannot criminalise expression. It does not mean private companies must amplify harmful content or be immune from liability for what they choose to promote. Newspapers are not immune from libel law, and they have editorial freedom. The same principle can apply to platforms that editorially curate content through algorithms. ### Aren't platforms just neutral hosts, not publishers? They were, once. But modern platforms use algorithms to decide what billions of people see, prioritising content that maximises engagement. That is editorial curation, not neutral hosting. If you are making editorial decisions about what to amplify, you bear editorial responsibility. ### Won't regulation just push harmful content to unregulated platforms? Some migration is inevitable, but the major platforms have network effects that make them hard to replace. Regulating Meta, X, TikTok and YouTube would cover the vast majority of social media use. Smaller platforms can be addressed as they grow. The alternative—doing nothing because you cannot regulate everything—is a counsel of despair. ## Sources - [UK Government — Online Safety Act 2023](https://www.gov.uk/government/collections/online-safety-act) - [Ofcom — Online safety regulation](https://www.ofcom.org.uk/online-safety) - [The Wall Street Journal — The Facebook Files (2021)](https://www.wsj.com/articles/the-facebook-files-11631713039) --- Daily Junction — https://dailyjunction.org/opinion/social-media-regulation-overdue