If AI is addictive, where does the responsibility lie – with big tech or its users?
Why Read This
What Makes This Article Worth Your Time
Summary
What This Article Is About
Researcher Bernd Stahl argues that generative AI — tools such as ChatGPT that produce text, images, and video — may be addictive, and that society urgently needs a framework for assigning responsibility for that harm. Although AI addiction is not yet formally recognised medically, Stahl cites growing evidence of neural patterns and behaviours consistent with addiction, including emotional dependency on chatbot companions, compulsive engagement, and the erosion of real-world social relationships. Drawing on the recent legal defeat of Meta and YouTube in a landmark social media addiction trial, he asks whether generative AI may be following the same trajectory.
Stahl identifies four groups of stakeholders who must collectively address AI addiction: governments and regulators, big tech companies, academic researchers, and civil society organisations. He draws historical parallels with tobacco and gambling — industries where corporate awareness of addictive properties was initially denied before being confronted through litigation and regulatory reform, including the WHO’s Framework Convention on Tobacco Control. His conclusion is that individual appeals to moderation are insufficient, and that structured, collaborative engagement between all parties is essential before AI becomes more deeply embedded in everyday life.
Key Points
Main Takeaways
AI Addiction: Emerging Evidence
Generative AI is not yet formally classified as addictive, but significant data already shows heavy use produces neural patterns and behaviours associated with addiction, including emotional dependency and compulsive engagement.
Social Media Sets a Precedent
Meta’s and YouTube’s landmark legal defeat in a social media addiction case signals that courts may increasingly hold tech companies accountable — a trajectory that could extend to generative AI providers.
The Tobacco Parallel
Like tobacco companies that knew smoking was addictive yet publicly denied it, AI companies may be aware of — and even exploiting — the addictive features of their products to increase user engagement and revenue.
Four Groups Must Act Together
Stahl identifies governments and regulators, big tech companies, academic researchers, and civil society organisations as the four stakeholder groups that must collaborate — none of them can solve the problem alone.
Big Tech Bears the Greatest Burden
AI companies hold user data needed to identify addictive features and stand to benefit financially from increased engagement — making them, in the author’s view, the most important party in addressing this issue.
Individual Moderation Is Not Enough
Appealing to users’ personal responsibility has consistently proved insufficient for other addictions — the article argues that structural solutions like age limits, labelling, and advertising restrictions are essential for AI too.
Master Reading Comprehension
Practice with 365 curated articles and 2,400+ questions across 9 RC types.
Article Analysis
Breaking Down the Elements
Main Idea
Generative AI Addiction Is a Shared Responsibility That Requires a Collective Response
Stahl’s central claim is that the addictive potential of generative AI cannot be left to any single actor — and that the current default of treating it as someone else’s problem is dangerous. By mapping the issue onto historical precedents with tobacco, gambling, and social media, he argues that only structured collaboration between governments, industry, researchers, and civil society can prevent AI addiction from becoming the next public health crisis.
Purpose
To Prompt Structured Public Debate About Who Is Responsible for AI Overuse
Stahl writes to move the conversation about generative AI from enthusiasm and alarm to a sober policy discussion about responsibility. His aim is to convince readers — and, implicitly, policymakers and industry leaders — that the moment to act is now, before AI becomes so embedded in daily life that harmful patterns become normalised and difficult to reverse.
Structure
Personal Hook → Problem Definition → Historical Analogies → Stakeholder Framework → Policy Prescription
The article opens with a personal anecdote about the author’s son using ChatGPT reflexively, then widens to define the problem of AI addiction and the question of responsibility. It draws historical analogies with tobacco and gambling before mapping four groups of stakeholders. It closes with a policy prescription — structured collaboration backed by evidence — modelled on the WHO’s tobacco treaty framework.
Tone
Measured, Cautionary & Policy-Oriented
Stahl’s tone is careful and academic without being detached — he acknowledges that the science is not yet settled while still making a clear, urgent argument. He avoids alarmism but conveys a sense of growing concern. The use of “I believe” and the personal opening anecdote give the piece a first-person accessibility typical of The Conversation, where scholars write for a general public audience.
Key Terms
Vocabulary from the Article
Click each card to reveal the definition
Build your vocabulary systematically
Each article in our course includes 8-12 vocabulary words with contextual usage.
Tough Words
Challenging Vocabulary
Tap each card to flip and see the definition
To find out or determine something with certainty through investigation or inquiry; to establish a fact through careful examination of available evidence.
“These companies own and have access to user data, which would be needed to ascertain the features that support or alleviate addiction.”
To make a problem, pain, or difficulty less severe; to partially relieve a harmful condition without necessarily eliminating it entirely.
“…user data, which would be needed to ascertain the features that support or alleviate addiction.”
Came to light; became known or evident, especially when referring to the revelation of facts that were previously concealed or unknown to the public.
“It eventually transpired that not only was smoking addictive and bad for your health, but that tobacco companies knew this.”
Resulting from or relating to an irresistible urge to behave in a certain way, especially against one’s conscious wishes; driven by impulse rather than rational choice.
“Much-discussed examples include emotional dependency on chatbot companions, compulsive engagement with them…”
More than enough; plentiful and sufficient in quantity or scope to meet a need — used in the article to indicate that historical precedents are numerous and well-established.
“…there is ample precedent showing how greater engagement from those involved with the issue may be achieved.”
Relating to the structure of human populations, defined by characteristics such as age, gender, income, or education; a demographic group shares particular statistical traits.
“The use of generative AI tools has exploded across different demographic groups.”
Reading Comprehension
Test Your Understanding
5 questions covering different RC question types
1The article states that generative AI is formally recognised by the medical community as an addictive substance or behaviour.
2According to the article, why does Stahl consider big tech companies the most important group in addressing AI addiction?
3Which sentence best explains why the article draws parallels between generative AI and the tobacco industry?
4Evaluate the following statements about the stakeholder groups identified in the article:
The article identifies four groups that must address AI addiction: governments and regulators, big tech companies, academic researchers, and civil society organisations.
According to the article, individual users bear no responsibility whatsoever for managing their own AI use, as overuse is entirely the fault of tech companies.
The article argues that appeals to individual moderation alone have proven insufficient for other addictions and are therefore unlikely to be sufficient for AI either.
Select True or False for all three statements, then click “Check Answers”
5The author closes by saying “the choices we now make will determine what acceptable use looks like for years to come.” What does this most likely imply about the author’s view of the current moment?
FAQ
Frequently Asked Questions
The article cites data showing that heavy use of chatbots and other generative AI systems produces neural patterns and behaviours associated with addiction. Specific examples mentioned include emotional dependency on chatbot companions, compulsive engagement, and the loss of real-world social relationships. The author also notes his own research team has published a paper arguing there is strong evidence for generative AI having addictive properties, though the medical establishment has not yet formally classified it as an addiction.
The article refers to a legal case in which Meta (parent company of Facebook and Instagram) and YouTube suffered a legal defeat in a trial focused on social media addiction. While the article does not provide detailed facts about the case, it uses this precedent to argue that courts are beginning to hold social media companies accountable for the addictive properties of their platforms — and that similar accountability may eventually apply to generative AI providers.
The WHO’s Framework Convention on Tobacco Control is an international treaty that brought together governments, public health organisations, researchers, and civil society to assess evidence on tobacco harm and create common rules — including restrictions on advertising, packaging requirements, and age limits. The article cites it as a model for how the global community can build structured, evidence-based consensus to regulate an addictive product, and suggests a similar multi-stakeholder framework may be needed for generative AI.
Readlite provides curated articles with comprehensive analysis including summaries, key points, vocabulary building, and practice questions across 9 different RC question types. Our Ultimate Reading Course offers 365 articles with 2,400+ questions to systematically improve your reading comprehension skills.
This article is rated Intermediate. It is shorter and more accessible than many Readlite articles, written in the clear, public-facing style of The Conversation. The argument is logical and well-structured, making it suitable for readers developing their comprehension skills. The challenge lies in tracking nuanced distinctions — such as the difference between formal medical classification and existing evidence, or between user responsibility and corporate responsibility — that are tested directly in the quiz.
Bernd Stahl is an academic researcher whose work focuses on the ethics and social implications of information technology and AI. He notes in the article that he is part of a research team that has published on generative AI addiction. The Conversation specifically publishes articles by academics writing for a general audience, which gives Stahl’s piece credibility as a policy-oriented piece grounded in active research — rather than journalism or opinion writing without academic backing.
The Ultimate Reading Course covers 9 RC question types: Multiple Choice, True/False, Multi-Statement T/F, Text Highlight, Fill in the Blanks, Matching, Sequencing, Error Spotting, and Short Answer. This comprehensive coverage prepares you for any reading comprehension format you might encounter.
