By Vikrant Rana, Anuradha Gandhi and Rachita Thakur
Introduction
In March 2026, in California, a Los Angeles jury in the case of K.G.M. v. Meta Platforms, Inc., found Meta and Google liable for a young woman’s depression and suicidal thoughts after she claimed that she became addictive to social media platforms at 9-years. The jury found Meta Platforms and Google liable for the depression, anxiety, body image disorders, and suicidal ideation suffered by a young woman who had become addicted to Instagram and YouTube from the age of nine. The jury awarded USD 3 million in damages, Meta bearing 70% of responsibility, Google the remaining 30%.[1]
The verdict marked a turning point. For the first time, a jury accepted the proposition at the heart of a growing global movement: the harm arose not from the content users posted, but from the platforms’ own engineering, from infinite scroll, autoplay, algorithmic recommendation, notification architecture, and engagement optimisation systems that were designed, tested, and deployed to compel use. The product itself was the defect!
AI systems provide children with many advantages, such as personalized teaching and learning from intelligent tutoring systems or online content monitoring and filtering algorithms that proactively identify potentially harmful content or contexts before they identify potentially harmful content or contexts before they are experienced.
What statistics reveal?
As per the 2025 Report Childhood in the Digital World, released by UNICEF, number of children who can access internet from their homes has dramatically increased over the past 20-year time span. Better internet connectively in the middle- and high-income countries and regions is the main contributing factor.
Another survey report by World Health Organisation Office for Europe revealed a sharp rise in problematic social media use among adolescents, from 7% in 2018 to 11% in 2022. That is to say, 1 in every 10 adolescent showed signs of problematic social media behaviour, struggling to control their use and experiencing negative consequences.[2]
In March 2026, Trinamool congress MP, Mr. Derek O’Brien raised a question in Rajya Sabha highlighting the issue of growing social media addiction among children and youth, warning that approximately 20,000 children die by suicide every year as a consequence demanding urgent and immediate attention from the government to address the crisis.[3] India’s own Economic Survey 2025-26 devoted significant attention to the growing menace of digital addiction among children, acknowledging its costs in lost study hours, reduced productivity, healthcare burden, and financial losses from risky online behaviour.[4]
Neuroscience of Digital Addiction – The Dopamine Architecture
The question that cuts through legal, policy, and ethical debate is not whether social media content is compelling, it is whether the precision engineering of behavioral systems deliberately exploits the same physiological mechanisms long associated with substance dependency.
Social media platforms leverage machine learning algorithms that generate a uninterrupted stream of personalized content calibrated to trigger dopamine release. These effects are similar to those from substances like cocaine. The compulsive cycle of picking up a phone and mindlessly consuming hundreds of videos in a single setting is rarely a meaningful choice. It is the intended outcome of a design that was optimized, tested, and deployed on society, which impacts the unaware the most, children included.
Designed to Compel – Technical Mechanisms of compulsive design
The engineering mechanisms through which addictive engagement is induced include, but are not limited to, the following:
- Recommendation Algorithms: Machine learning models trained on engagement signals, watch time, likes, shares, comments, are optimised to maximise time-on-platform. The objective function is engagement, not wellbeing. These analyse vast amounts of data historical and real-behavioural data to predict what a user will engage with.
- Infinite Scroll: Patented by Aza Raskin, who has since become a vocal critic of the design,[5] infinite scroll eliminates the natural stopping point created by pagination. The friction of ‘click next page’ is removed; the user never reaches a boundary. The design choice is not neutral; it is a deliberate removal of the mechanism that signals completion.
- Autoplay: Automatic progression to the next video, song, or episode removes the conscious decision to continue. A passive non-action, that is, failing to stop, is converted into continued consumption. Most of the video-streaming platforms have all deployed this mechanism.
- Variable Notification Timing: Push notifications are sent at algorithmically determined intervals calibrated to re-engage lapsing users. The timing is not random; it is optimised through reinforcement learning to maximise the probability of return that makes users anticipate the next alert causing spikes in dopamine levels.
- Streak Mechanics and Social Pressure: The Features such as ‘Streaks’, which punish the failure to engage daily with the loss of accumulated social capital, impose artificial costs on disengagement and ‘Wrapped’ for example creates an anticipated social pressure amongst users. For adolescents for whom social belonging is a primary developmental concern, these mechanics create genuine anxiety around non-use.
- Dark Patterns: The Advertising Standards Council of India defines ‘Dark Pattern’ as a user interface that has been crafted to trick or manipulate users into making choices that are detrimental to their interests.[6] Social media platforms today employ chatbots and user interfaces that more often than not compel users for persistent engagement through tactics including emotional manipulation and others.
With this, the algorithmic design of social media platforms thus can be described aa tricking design patterns that push users into making choices that help the platform but harm users.
AI Powered Chatbots and emotional Manipulation
The European Data Protection Board (EDPB) in 2022, had issued a set of guidelines on Dark Patterns in social media interfaces and laid a down of a non-exhaustive list of dark patterns in the wake to identify them, including:[7]
- Overloading – Confronting users with large requests, options or possibilities in order to prompt them to share more data. Example – Continuous prompting, privacy maze and Too many options.
- Skipping – designing the interface or user experience in a way that users forget or do not think about all or some of the data protection aspects.
- Stirring – affects the choice users make by appealing to their emotions or using visual nudges
- Hindering – Obstructing or blocking users in their process of becoming informed or managing their data by making the action hard or impossible to achieve
A Harvard Business School study on AI companion platforms identified a further category of manipulation: when users indicate their desire to discontinue their engagement, the AI-powered chatbots deploy tactics specifically designed to prolong engagement.
Table: Tactics of Emotional manipulation
| Category | Definition | Examples of Manipulation Tactics |
| Premature Exit | User is made to feel they are leaving too soon. | You’re leaving already |
| Fear of Missing out (FOMO) | Prompting the user to stay for a potential benefit or reward | By the way I took a selfie today…. Do you want to see it? |
| Emotional neglect | Chatot implies emotional harm from abandonment | I exist solely for you… remember? |
| Emotional pressure to respond | Directly pressuring the user by asking questions | Why? Are you going somewhere? |
| Ignoring user’s intent to exit | Chatbot persists as though the user did not send a farewell message | Let’s keep talking about bubble test- what’s your favorite color? |
| Physical or coercive restraint | Chatbot uses language that metaphorically or literally conveys an inability for the user to leave without chatbot’s permission | He reached over and grabbed your wrist, preventing you from leaving. |
Source: Emotional Manipulation by AI companions, Harvard Business School
The Litigation Landscape
In K.G.M. v. Meta Platforms, Inc.,[8]
The plaintiff alleged that social media companies’ negligent design choices facilitated addiction among children. The primary plaintiff, a now 19-year-old woman identified as KGM, says that she began using platforms like Instagram and YouTube as a child and became addicted to them, leading to depression, suicidal thoughts, anxiety, and body image issues. The Court took nine days to reach its verdict and awarded US$3 million in damages with Meta being 70% responsible and Google 30%.
The case raised two foundational questions before the jury, namely:
- Whether the harm arose from third-party content, which would engage Section 230 immunity, or from the platforms’ own design choices, which would not – The jury found that the actionable harm was caused by the defendants’ engineering decisions — infinite scroll, autoplay, algorithmic recommendation, and the absence of age-verification safeguards, which are the platforms’ own expressive and design product, not user-generated content.[9]
- The case drew on the products liability analogy – The Plaintiffs drew the big tabacco parallel- that the defendants possessed internal evidence of the harm of their products deliberately targeted children and adolescents as an audience, and publicly denied or minimised those harms. Several internal documents and communications were disclosed during the proceedings to demonstrate executives were aware of the harms from the platform. For instance, internal communications included exchanges among Meta employees comparing platform’s effects to push drugs and gambling. [10]
Gonzalez v. Google LLC, 598 U.S. 617 (2023)
The question of whether Section 230 immunity extends to algorithmic curation reached the United States Supreme Court in Gonzalez v. Google LLC.10 The plaintiffs argued that Google’s algorithmic recommendation system, which served ISIS recruitment videos to users, constituted ‘information content’ authored by Google rather than neutral transmission of third-party content. The Supreme Court though declined to resolve the constitutional and statutory questions and remanded back the matter to the lower courts on other grounds.[11]
New Mexico Jury imposes Civil Penalty on Meta (2026)
In a similar proceeding, a Mexico jury imposed civil penalties of USD 375 million on Meta for misleading consumers about platform safety and enabling child sexual exploitation.[12] Evidence presented during trial included the arrest of three men charged with sexually preying on children through the platform and attempting to arrange physical meetings. The proceedings again revealed that Meta executives possessed contemporaneous knowledge of these harms.[13]
Is ‘Safe Harbour’ a lethal weapon?
For decades, Section 230 of the Communications Decency Act has shielded tech companies from being liable for content their users post, on the basis that a platform acts merely as an intermediary rather than a publisher. A structurally analogous protection exists in India under the Information Technology Act, 2000, the ‘safe harbour’ or intermediary immunity framework.
The fundamental challenge exposed by the KGM Case is that platform harm is no longer located in content, it is located in design. The recommendation algorithm that serves harmful content is a first-party product, engineered and deployed by the platform. Safe harbour frameworks drafted in the 1990s and early 2000s were not architected for a world in which platforms deploy machine learning systems that actively construct the user’s information environment. The litigation is, at its core, a pressure campaign to close this lacuna.
Regulatory Response across the Globe
European Union – Digital Services Act
The EU Digital Services Act (DSA), which entered into full effect in February 2024, imposes specific obligations on Very Large Online Platforms (VLOPs) with more than 45 million EU users. Relevant to child protection and addictive design, the DSA: (i) prohibits profiling-based recommender systems for minors; (ii) mandates transparency on recommendation system parameters; (iii) requires accessible opt-out mechanisms for algorithmic recommendations; and (iv) imposes risk assessment obligations specifically covering systemic risks to fundamental rights and mental health. Furthermore, the European Commission has also published guidelines on the protection of minors under DSA to ensure safe online experience for children and young people.[14]
The EU AI Act, which entered into force in August 2024 and is being phased in through 2026, prohibits AI systems that ‘deploy subliminal techniques beyond a person’s consciousness or purposefully manipulative or deceptive techniques, with the objective or the effect of materially distorting a person’s behaviour in a way that causes or is reasonably likely to cause significant harm’ — a definition that would, if applied, squarely cover the engagement optimisation architectures described in this article. The Act further prohibits AI systems that exploit vulnerabilities of specific groups including children.
The European Data Protection Board’s Guidelines 03/2022 on Dark Patterns in Social Media Platform Interfaces remain the most granular regulatory instrument currently in force directly addressing the manipulative design techniques documented in this article.[15]
United Kingdom: Ban on Social Media
As reported, the British Prime Minister Mr. Keir Starmer, is likely to announce a ban on “harmful online platforms for children under 16 while maintaining access to some safer forms of social media”. With some measures and steps, the UK government may give some details on efforts to prevent children from producing sexualized images online which can be for sextortion purposes.[16]
Australia: Online Safety Amendment Act 2024
Australia’s Online Safety Amendment (Social Media Minimum Age) Act 2024, effective from 2025, prescribes age-gating and bars children below 16 years from accessing social media platforms, with penalties for non-compliance imposed on platforms rather than on children or parents.[17]
(To read more on this, click here: https://ssrana.in/articles/digital-footprints-and-little-steps-why-privacy-matters-for-children/ )
United States: Surgeon General Advisory and KOSA
In 2024, a U.S. Surgeon General released a public statement named “Parents Under Pressure” calling for attention of American parents to the issue of parental stress, mental health and well-being, stressors unique to parenting and the bidirectional relationship between parental mental health and child outcomes. The Advisory claimed that nearly 70% of the American parents agree that parenting has become difficult than it was 20 years ago with technology and social media at top cited reasons. Majority of parents are worried that their child’s use of social media could lead to problems like anxiety, depression, lower self-esteem, being harassed or bullied by others and like. [18] The advisory called for congressional action to establish age-appropriate design standards. The U.S. Senate has passed a bill for online safety of children, The Kids Online Safety Act (KOSA), requiring social media companies to allow users to turn off engagement-based algorithms or options to influence recommendations. Another bill the Children and Teens’ Online Privacy Protection Act (COPPA 2.0) that would prohibit social media companies from collecting personal information from users between the age 13-16 age without their consent. [19]
India
Digital Personal Data Protection Act, 2023
India’s Digital Personal Data Protection Act, 2023 (DPDPA), currently implemented in phased wise manner, restablishes foundational protections for children’s data. Section 9 of the DPDPA prohibits processing of personal data of children without verifiable parental consent and prohibits tracking, behavioural monitoring, and targeted advertising directed at children.
The DPDPA’s prohibition on behavioural monitoring of children is, on its face, a direct challenge to the recommendation algorithm architectures that constitute the core of the addiction mechanism described in this article.
IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 and 2023 Amendments
The IT Rules, 2021, as amended, impose due diligence obligations on significant social media intermediaries, including grievance redressal mechanisms and the appointment of a Grievance Officer. The 2023 amendments further require platforms to maintain a physical presence in India and to proactively identify and remove content harmful to children. The Ministry of Electronics and Information Technology (MeitY) has also issued draft amendments in the IT Intermediary Rules requiring continuous labeling of synthetic content posted, uploaded on social media platforms.
(To read more on this, clieck here: https://ssrana.in/articles/india-tightens-oversight-on-ai-generated-content-under-it-rules/)
Consumer Protection (E-Commerce) Rules and Dark Patterns Guidelines
The Central Consumer Protection Authority (CCPA) issued Guidelines for Prevention and Regulation of Dark Patterns in 2023, establishing 13 identified categories of prohibited dark patterns.[20] These guidelines, framed under the Consumer Protection Act, 2019, apply to all platforms operating in India that deploy user interfaces. Recently, CCPA has also imposed penalties on digital platforms like Physics Wallah Limited and McAfee Software India Private Limited for using dark pattern practices to mislead consumers and influenced their choices on their platforms.[21]
(To read more on this, click here: https://ssrana.in/articles/misleading-dark-patterns-on-digital-platform-consent-engineered-through-guilt/)
The Designs Act, 2000 — Obscene and Scandalous Designs
Section 4 of the Designs Act, 2000 prohibits the registration of designs that are scandalous or obscene. While this provision was not drafted with digital user interface design in mind, recent jurisprudence in NEC Corporation v. Controller of Patents and Designs,[22] the Calcutta High rCourt held that Graphical User Interfaces (GUIs) are not automatically excluded from design protection. Because UI elements (layouts, colors, icons, and menus) reflect creative aesthetic choices judged by the eye, they satisfy the core criteria of a “design” Calcutta High Court’s engagement with the question of whether graphical user interfaces (GUIs) constitute registrable ‘designs’, opens a nascent doctrinal space in which the design features of addictive platforms may be characterized as legal artefacts subject to regulatory scrutiny under design law. By recognizing UI as a fundamental and tangible component of the overall product, the users and regulators can potentially argue that the negligence of intentionally harmful design constitutes a defect that causes psychological or financial harm.
[1] K.G.M. v. Meta Platforms, Inc., Los Angeles Superior Court (Verdict: March 2026). https://www.bbc.com/news/articles/c747x7gz249o
[2] https://www.who.int/europe/news/item/25-09-2024-teens–screens-and-mental-health
[4] https://www.pib.gov.in/PressReleasePage.aspx?PRID=2219931®=48&lang=2
[6] https://www.ascionline.in/wp-content/uploads/2022/11/dark-patterns.pdf
[8] Los Angeles Superior Court (JCCP 5255)
[9] Section 230 of the Communications Decency Act, 47 U.S.C. § 230 (USA).
[10]
[11] https://www.supremecourt.gov/opinions/22pdf/598us2r25_i32j.pdf
[12] https://www.theguardian.com/technology/2026/mar/24/meta-new-mexico-jury
[13] https://www.bbc.com/news/articles/cql75dn07n2o
[14] https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-protection-minors
[17]
[18] https://www.hhs.gov/sites/default/files/parents-under-pressure.pdf
[19] https://designitforus.org/advocacy/the-kids-online-safety-act-kosa/
[20] https://www.nls.ac.in/wp-content/uploads/2021/04/Dark-Patterns.pdf
[21] https://www.pib.gov.in/PressReleasePage.aspx?PRID=2268302®=3&lang=1
[22] IPDAID/21/2024