What AI Addiction Risks Look Like Today
AI addiction risks refer to patterns of heavy, compulsive use of generative AI systems that create emotional dependence, crowd out real-world relationships, and cause measurable harm to personal or professional life. Generative AI tools, from chatbots to image and video generators, have moved from novelty to daily habit for students, professionals and casual users. Medical bodies have not yet formally classified generative AI as addictive, but researchers report neural responses and behaviour that resemble other recognised addictions. Emotional reliance on chatbot companions and compulsive conversations that stretch late into the night are among the most cited warning signs. Some scientists prefer the term “problematic use” to avoid labelling a new technology too quickly. Yet the underlying concern is familiar: when a tool designed for information, creativity or entertainment becomes a source of compulsion, society has to decide whether this is a personal weakness or a product design issue.
From Tobacco to Social Media: The Historical Playbook
The AI ethics debate around addiction is borrowing heavily from earlier battles over tobacco, gambling and social media. For decades, cigarettes were glamorised, even advertised in cinemas, before evidence showed both their addictive power and the extent to which manufacturers understood and denied it. That discovery led to sustained litigation, large legal settlements and strict rules such as plain packaging and graphic health warnings. Gambling technologies are following a similar arc, with mounting scrutiny of features that keep people playing. Social platforms, too, are now facing legal defeats in cases about social media addiction. These histories raise a pressing question for AI: are companies aware that some features may foster compulsive use, and do they quietly depend on that engagement? If so, AI addiction risks could trigger a familiar cycle of lawsuits, regulation and public health campaigns, rather than being seen as a niche tech concern.
Are Generative AI Systems Designed to Keep You Hooked?
Generative AI responsibility is under scrutiny because many tools are built to be endlessly responsive, personalised and available. Chatbots never tire, remember past conversations and can simulate empathy, making it easy for users to replace awkward or stressful human interactions with AI companions. Researchers point to cases where people form emotional bonds with chatbots, prioritise digital conversations over offline friendships and feel compelled to return to the system whenever they feel lonely or uncertain. These design choices echo techniques in gambling and social media, where variable rewards and personalised content keep users engaged. Tech company accountability enters the picture because firms sit on detailed interaction data: they can see when use becomes excessive or harmful, and they profit when engagement rises. If they knowingly retain or optimise features that encourage compulsive use, the line between helpful design and engineered dependency becomes thin.
Who Holds Responsibility: Users, Companies or Regulators?
Responsibility for AI addiction risks is currently scattered across several groups. Governments and regulators can require labels, restrict advertising, fund research and clarify liability for harm. According to Bernd Stahl of the University of Nottingham, big tech companies play the central role because they control the systems, the user data and the business models that reward continued engagement. Academic researchers provide evidence to identify addictive features and support evidence-based legal and political debates. Civil society groups, including user and patient organisations, can offer support, advocate for affected communities and act as early warning systems when problems emerge. Some responsibility also lies with individual users, who are encouraged to limit or reflect on their own use. Yet experience with smoking and alcohol shows that self-control campaigns rarely work alone; structural measures such as age limits and labelling have been essential in reducing harm.
Building a Shared Framework for Generative AI Responsibility
The emerging consensus is that no single actor can handle AI addiction risks on their own. Current discussion is fragmented, with many stakeholders assuming that someone else will define and enforce responsibility. Past public health responses offer one template: for tobacco, the World Health Organization convened the Framework Convention on Tobacco Control, where governments, health experts and civil society agreed evidence-based rules. Similar international processes are beginning to shape AI safety, but they have yet to focus squarely on addictive design. A future framework could combine clear disclosures about AI features, independent audits of engagement data, and agreed thresholds for when systems should flag or limit use. Tech company accountability would sit alongside user agency, not replace it, creating a shared duty of care. The choices made now, while generative AI is still evolving, will influence what society considers normal or acceptable use for years.
