Meta, the U.S. big tech company, settled lawsuits in August alleging that social networking services such as Facebook and Instagram harmed children and adolescents. The company agreed to pay a combined $18 billion to U.S. state governments and to immediately implement sweeping changes designed to protect children and teenagers from the harms of social media use. The case is read as a clear signal that citizens can fight back against big tech companies that inflict social harm.
The lawsuits appear to be only the opening act in a larger legal conflict over the social harms arising from the relentless pursuit of performance by major artificial intelligence companies. Those harms are likely to overlap heavily with the problems created by social media. AI that serves as an adviser, a romantic partner, a therapist, a friend or a financial consultant is likely to be optimized for continuous user engagement — that is, addiction — rather than for users' long-term well-being. We are already watching AI developers push to bring AI into schools. According to research by the Massachusetts Institute of Technology, this not only undermines critical thinking skills but is also likely to breed early dependence on AI, a model that looks increasingly similar to the early addiction playbook of tobacco companies.
On the question of harm to children, just as social media exposed children to content for which they were emotionally and psychologically unprepared, AI offers a platform that is extremely powerful yet can be only imperfectly regulated when it comes to illegal drugs, harmful social norms, extremist ideologies and self-harm. Just as social media made the exploitation of children possible through pornography and other channels, AI can do the same through its ability to produce harmful and illegal content. As a result, it is only a matter of time before children and adolescents become addicted to AI in the same way they became addicted to social media. That will lead to legal conflicts with much in common with the recent social media lawsuits.
Legal conflicts in other areas surrounding AI companies are also foreseeable. The first concerns the social harm arising from AI companies' unauthorized use of human-created content online, and more broadly the AI models and AI-powered robots that could replace human labor across a range of occupations. Large language models such as ChatGPT and Claude were trained on an enormous volume of human-created text content, from books to blogs and from encyclopedias to academic papers. AI image-generation software was trained not only on the work of the most influential photographers, artists and animators but also on ordinary photographs posted online. The original content creators received no compensation for the use of their work as training data, nor did they consent to that use.
From a researcher's point of view, ChatGPT, having absorbed all of this material, will soon be able to write papers and essays closely resembling a researcher's own output. In a sense, our human capital has been expropriated without consent and incorporated into AI models. Imagine, for instance, that you are an Uber driver and your job is to directly train the AI that will replace human drivers. Or imagine that you were one of Facebook's employees and your keystrokes were collected by your employer to train an AI that may ultimately replace you. The experience of having your skills taken by the very thing that seeks to eliminate your job goes beyond provoking anger — it will clearly lead to legal conflict.
The scale and growth rate of the AI sector mean this is not all. We may face lawsuits over environmental destruction, violations of labor standards across supply chains, or corruption in the process of signing data center contracts. This will affect not only AI developers but also the companies that produce AI infrastructure and hardware. Many of them are Korean firms, including SK hynix, Samsung Electronics, LS Electric, Hyundai Motor, SK Telecom and NAVER. That said, the growing dependence of countless companies and government agencies on AI technology also suggests these firms are too important to face very severe punishment. Academics may not be compensated as much as they think for having educated AI models. Still, as with the Meta lawsuits, public efforts against the harms of social media demonstrate the capacity to successfully fight the most serious harms of AI. Civil society, regulators, governments and companies should prepare for the battle to come.