Platform liability: is the Meta and Google conviction in the USA a turning point?

The week of March 24, 2026 will go down in the annals of digital law.

Within forty-eight hours, two American courts handed down decisions that could well reshape the boundaries of platform liability on a global scale.

In New Mexico, a civil court jury in Santa Fe ordered Meta to pay $375 million for violating the state's unfair business practices legislation by enabling the sexual exploitation of children on Facebook and Instagram. The following day, in Los Angeles, a second jury awarded $6 million in damages (70% from Meta, 30% from YouTube) to a young Californian woman who became addicted to social media as a teenager.

The fundamental distinguishing feature of these decisions lies in their angle of attack: it is not the hosted content that is targeted, but the very design of the platforms and their structural failings in protecting minors.

The content shown to users is now essentially determined by algorithms, such that the supposedly 'passive' role of the host, which was the basis of the limitation of liability, simply no longer exists in practice.

The plaintiffs' lawyers deliberately circumvented the historic protection of Section 230 of the Communications Decency Act - the American equivalent, in spirit, of our conditional immunity regime for hosts under the LCEN - by shifting the debate from the field of content to that of design defects. A strategy borrowed directly from the tobacco industry litigation.

These decisions come at a time when Europe is keeping pace. On February 6, 2026, the European Commission published its preliminary findings that TikTok infringes the DSA due to its addictive design. In France, the parliamentary inquiry committee report on the psychological effects of TikTok on minors (report no. 1770, filed September 4, 2025, chaired by Arthur Delaporte with Laure Miller as rapporteur), formulated 56 recommendations, some of which challenge the very architecture of the intermediary liability regime.

On March 26, 2026, the Ministry of National Education filed a report with the Paris public prosecutor's office under Article 40 of the Code of Criminal Procedure, targeting TikTok's practices.

The convergence of these events calls for fundamental reflection: is the liability regime for digital platforms, as it was designed in Europe at the beginning of the 2000s, still suited to the reality of intermediaries whose business model is precisely based on capturing attention, particularly that of the youngest users?

I. The Classic Regime of the Passive Host: A Paradigm in Crisis.

A. The Original Architecture: the LCEN and the Fiction of Passivity.

Since the e-Commerce Directive of June 8, 2000 (Directive 2000/31/EC), transposed into French law by Law No. 2004-575 of June 21, 2004, on Confidence in the Digital Economy (LCEN), European and French law have rested on a cardinal distinction: the host, defined as any person providing storage of content supplied by third parties, can only be held liable for the content it hosts if it has actual knowledge of its unlawful nature and has not acted promptly to remove it.

The LCEN enshrined a principle of conditional immunity, coupled with a prohibition on any general monitoring obligation.

This regime was based on an economic and technical premise that was coherent in the early 2000s: the host was conceived as a technical service provider, a neutral conduit through which information passed over which it exercised no editorial control.

In 2004, Facebook had just been created and TikTok did not exist. The Constitutional Council, in its decision No. 2004-496 DC of June 10, 2004, had moreover issued an interpretive reservation specifying that host liability required manifestly unlawful content or the non-execution of a court order.

This fiction of passivity has been progressively eroded by the reality of contemporary platforms.

The CJEU, in the Google France judgment of March 23, 2010, then in the L'Oréal v. eBay judgment of July 12, 2011 and the YouTube and Cyando judgment of June 22, 2021, refined the boundary between active and passive roles: when an intermediary exercises control over content, selects it, organizes it, promotes it, it falls outside the protective host regime.

But this case law, built on a case-by-case basis, has never truly drawn all the consequences of what it means for a recommendation algorithm to propose specific content to each user based on their profile, age, and vulnerabilities.

B. The DSA: A Reform That Did Not Change the Paradigm.

Regulation (EU) 2022/2065 of October 19, 2022 on Digital Services (DSA), which became fully applicable on February 17, 2024, did not fundamentally challenge this architecture. Article 6 of the DSA reiterates the principle of conditional host immunity: the host is not liable for stored content provided it has no knowledge of its unlawful nature or acts promptly once it gains such knowledge.

Article 8 confirms the absence of a general monitoring obligation, stating that intermediary service providers are not subject to any obligation to monitor the information they transmit or store, nor to actively seek facts or circumstances indicating illegal activities.

The DSA has certainly introduced new obligations for very large online platforms (VLOPs): systemic risk assessments (Articles 34 and 35), algorithmic transparency, independent auditing.

However, the fundamental question remains: when a recommendation algorithm serves content related to self-harm, eating disorders or suicide to a 13-year-old minor, not because of an active choice by the user but solely through the personalized recommendation system, can we still speak of a passive role for the platform?

The parliamentary report notes that dangerous content spirals form independently of the user's active choices, with mere video viewing sufficient to trigger the algorithmic spiral. The old social media models required users to choose content and subscribe to accounts; TikTok imposes a feed determined by the algorithm, with no real margin of control.

II. The American Turning Point of March 2026: Design as the Triggering Event.

A. The New Mexico Verdict: $375 Million for Systemic Failure.

On March 24, 2026, after less than a day of deliberation and six weeks of proceedings, a jury at the Santa Fe civil court found that Meta had committed 37,500 willful violations of New Mexico's Unfair Practices Act, each carrying a civil penalty of $5,000, for a total of $375 million.

Attorney General Raúl Torrez had filed suit against Meta at the end of 2023 following an undercover investigation: agents had posed as children on Instagram and Facebook, documenting the sexual solicitations received and the inadequacy of Meta's responses.

The prosecution's argument rested on two pillars. On one hand, Meta had communicated misleadingly about its child protection measures. During closing arguments, Linda Singer, representing the prosecution, accused Meta of encouraging excessive use of its platforms by minors while being fully aware of the associated risks.

On the other hand, internal documents revealed during the trial demonstrated that the company was perfectly aware, as early as 2015, of the massive presence of minors on its platforms: a 2018 internal document estimated that four million Instagram accounts belonged to children under 13, and that approximately 30% of 10-12 year-olds used the platform in the United States. In December 2015, Mark Zuckerberg himself had set a target to increase the average time spent on Instagram by 12% over three years.

The verdict was immediately hailed as a historic victory by the Attorney General, who emphasized that the sum sent a clear message to the entire technology sector. Meta announced it would appeal. Phase 2 of the trial, scheduled for May 2026, will address whether Meta constitutes a public nuisance and whether the court can order structural modifications to its platforms, including effective age verification and predator removal.

B. The Los Angeles Trial: The Addictive Design of the Platform.

On March 25, a jury at the Los Angeles Superior Court awarded $6 million in damages ($3 million compensatory and $3 million punitive) to a young Californian woman as part of multidistrict litigation (MDL 3047) consolidating more than 2,100 federal lawsuits against Meta, TikTok, YouTube and Snapchat.

The plaintiff, identified by the initials K.G.M. (first name: Kaley), aged 20 at the time of trial, had stated that she developed an addiction to YouTube from the age of 6, then signed up for Instagram as a child, circumventing her mother's parental controls. She testified to spending entire days on social media and systematically using filters to alter her photos, developing severe depression, anxiety, body dysmorphia and self-harm.

What fundamentally distinguishes this litigation from previous cases is the legal strategy: the lawyers do not argue on the basis of content, but on that of design defect. To achieve a conviction, the plaintiff's lawyers had to establish that the platforms' negligence in the design of their services constituted a substantial factor in the harm caused.

The jurors did not have to rule on the content viewed by Kaley, as the companies were protected in this regard by Section 230. Ten of the twelve jurors found both companies negligent, holding Meta liable for 70% and YouTube for 30%. This is a direct transposition of the methodology that, in the 1990s, made it possible to establish the liability of the tobacco industry.

It is significant that TikTok (ByteDance) and Snap Inc. (Snapchat), initially included in this bellwether trial, chose to settle out of court before the proceedings began, for undisclosed amounts. This choice reveals a legal risk assessment high enough to prefer a confidential financial settlement.

The trial was marked by Mark Zuckerberg's sworn testimony for over six hours on February 18, 2026. Confronted with his own internal documents, Meta's CEO acknowledged being slow to implement controls on minors' use of his apps. Instagram only began asking users for their date of birth in 2019. Crucially, a 2023 internal document revealed that only 1.1% of teenagers used the 'daily time limit' feature and 2% used 'take a break', undermining Meta's argument that parental control tools were sufficient.

III. The European and French Response: Between the DSA and National Initiatives.

A. The European Commission's Preliminary Findings Against TikTok.

On February 6, 2026, the European Commission published its preliminary findings as part of the formal proceedings opened on February 19, 2024 against TikTok under the DSA (press release IP/26/312).

The Commission considers that TikTok infringes the regulation due to its addictive design, which relies on features such as infinite scrolling, auto-play of videos, push notifications and a highly personalized recommendation system. According to the Commission, these features fuel the desire to keep scrolling and place users' brains on autopilot, fostering compulsive behavior and reducing their self-control.

The Commission further criticizes TikTok for failing to adequately assess how these features could harm the physical and mental well-being of its users, including minors and vulnerable adults.

It considers that TikTok should modify the fundamental design of its service.

ARCOM, as the Digital Services Coordinator for France, welcomed its contribution to this investigation. Its president, Martin Ajdari, recalled that platforms have an obligation to protect their users, and that these findings constituted a clear warning to all platforms that might attempt to evade this responsibility. If confirmed, TikTok could face a fine of up to 6% of its annual worldwide turnover.

The major interest of this proceeding is that it uses the DSA not on the classic ground of content moderation, but on that of the very design of the service. For the first time at European level, it is no longer just illegal content that is targeted, but the platform's architecture and its influence on human behavior. Commission Vice-President Henna Virkkunen emphasized that social media addiction could no longer be an accepted business model at the expense of children's health.

B. The Parliamentary Report on TikTok's Danger to Minors: A Damning Assessment, Decisive Recommendations.

The inquiry committee's report (report no. 1770, 273 pages, adopted unanimously) paints an unequivocal picture.

In the foreword, Chairman Arthur Delaporte recounts the chilling testimonies heard during the closed-door hearings of May 15, 2025, in which families gathered behind the Algos Victima collective described the suicide attempts, suicides, depression, and eating disorders of their children, encouraged by an algorithm that pushes toward the worst, that locks them in a bubble from which adults are logically excluded.

The rapporteur, Laure Miller, concludes: this platform knowingly exposes our children to toxic, dangerous, addictive content, too often evades the rules, shirks its responsibilities, and thrives in an attention economy that pretends to ignore the psychological and social devastation it causes.

The figures are telling: 72% of 16-18 year-olds and 64% of 12-15 year-olds use TikTok at least once a day. The average age of first access to social media is 8.5 years. The report identifies several TikTok-specific features: advertising exposure reaching one in three videos, much shorter and potentially more viral content, and above all a particularly confining algorithm that is difficult to correct or force open to other interests.

The recommendation system leads to themes related to young people's psychological vulnerabilities, drawing them into spirals of content related to depression, suicide, self-harm or eating disorders.

Among the report's 56 recommendations, several are particularly significant for lawyers. Recommendation No. 10 advocates commissioning legal experts to assess the merits of evolving the liability of online social network services toward publisher status - which would remove them from the protective host regime.

Recommendation No. 11 proposes supporting, within the framework of the future Digital Fairness Act, the inclusion of provisions strengthening obligations to diversify recommended content and imposing a mandatory element of randomness in algorithms. Recommendation No. 12 aims to introduce into European law an obligation of algorithmic pluralism, inspired by the principle of media pluralism enshrined in Article 34 of the Constitution.

The committee chairman, Arthur Delaporte, expresses in his foreword a more radical position than the rapporteur's: for him, banning social media access for under-15s amounts to admitting that we have given up on regulating digital giants. His conviction, shared during our exchanges, is that the law should compel platforms to become proactive in moderating unlawful content, and allow their liability to be engaged if they fail to do so. This is no longer a question of content: it is a question of design.

In parallel, the National Assembly adopted on first reading, on January 26, 2026, the bill aimed at protecting minors from the risks posed by social media use, with a provision banning access for under-15s.

IV. Toward a New Paradigm: Liability for Defective Design.

A. Moving Beyond the Host/Publisher Distinction.

The fundamental lesson of these converging developments is that the classic distinction between host and publisher, which has structured our law for twenty years, no longer reflects reality. A social network like TikTok, Instagram or YouTube is neither a passive host that merely stores content provided by third parties, nor a publisher in the traditional sense that chooses to publish specific content.

It is an intermediary of a new type whose business model is based on the algorithmic curation of attention. The parliamentary report states bluntly: platforms take refuge more than is justified behind their questionable host status.

When a recommendation algorithm selects, from among millions of videos, those that will be presented to a 13-year-old user based on their psychological profile, viewing history, reaction time, and pauses - and when this selection has the measured effect of drawing them into spirals of content related to suicide or self-harm - one cannot reasonably argue that the platform plays a purely passive role.

Amnesty International's 2023 report demonstrated, through real-world tests, that simply creating fake accounts for 13-year-old minors and scrolling through the For You feed for several hours was enough to generate a concentration of content related to psychological distress, without any active interaction.

The French parliamentary report confirmed this phenomenon. It is this factual finding that led recommendation No. 10 of the report to advocate assessing the merits of evolving the status of platforms toward that of publisher.

B. The Imperative of a Proactivity Obligation.

The heart of the debate is now the obligation of proactivity.

The current regime, under both the LCEN and the DSA, only imposes a reactive obligation on platforms: removing content after notification of its unlawfulness. But when the harm results not from specific content but from the very design of the service, a reactive obligation is structurally insufficient. Unlawful content is a symptom; addictive design is the cause.

The DSA began this shift by imposing on very large platforms an obligation to assess systemic risks (Article 34) and to mitigate those risks (Article 35). But these obligations are framed in terms of means, not results. The parliamentary report goes further by recommending that platforms be compelled to become proactive in moderating unlawful content, and that their liability be engaged if they fail to do so.

Recommendations Nos. 10, 11 and 12 outline a framework in which platforms would be required not only to moderate content, but to rethink their algorithmic architecture.

However, the risks of such a change must be weighed. The prohibition on general monitoring obligations, enshrined in Article 8 of the DSA, is not a legal anachronism: it protects freedom of expression and prevents platforms from being turned into private censors.

The Court of Cassation recalled, in a judgment of March 27, 2024 (Cass. com., March 27, 2024, No. 22-21.586), that hosts cannot be compelled to implement preventive and unlimited blocking of hosted content.

Conclusion.

A consensus is forming around a simple idea with considerable legal consequences: when a platform designs a service whose recommendation algorithm exposes minors to content dangerous to their mental health, liability cannot be limited to a duty to remove content after notification. It must extend upstream, to the very design of the service itself.

The LCEN regime, designed in 2004 for an internet where hosts were mere technical service providers, and the DSA, which updated this regime without changing its philosophy, are no longer equal to the challenge.

The platforms of 2026 are not the hosts of 2004. They are architects of attention, and their algorithmic design has a measurable impact on the mental health of millions of minors. The law must draw the consequences.

Building a genuine law of responsible digital design is now one of the great legal undertakings of the coming years. It will need to reconcile three requirements: the effective protection of minors, which can no longer remain mere wishful thinking; respect for freedom of expression and the refusal to turn platforms into private censors; and the proportionality of imposed obligations, so as not to stifle innovation.

The balance is difficult, but recent developments show that the status quo is no longer tenable. Platforms that refuse to rethink their design will henceforth face a liability that neither Section 230, nor Article 6 of the DSA, nor the vestiges of the LCEN will suffice to dismiss.