In conversation with Yasmine Rejeb

Yasmine Rejeb is a researcher, policy analyst, and author working at the intersection of artificial intelligence, global governance, and international law. A former Global Fellow at Columbia University’s Committee on Global Thought, Yasmine contributed to the Youth in a Changing World project, a large-scale qualitative study involving more than 400 young people across 10 countries. Through this research, she examined how young people navigate rapid technological, social, and economic change, offering valuable insights into the opportunities and challenges that AI presents for education, employment, mental health, and autonomy. Trained in both law and international relations, she holds a Master’s degree in International Economic Law from Paris 1 Panthéon-Sorbonne, where her research focused on the regulation of artificial intelligence in financial services. Her work combines legal analysis with a broader interest in how technology influences human development, democratic societies, and the future of opportunity.

In this conversation, we explore how young people are actually using conversational AI, not only as a tool for learning and productivity, but increasingly as a source of advice, emotional support, and self-reflection. We discuss the implications of these changes for autonomy, identity, education, and the future relationship between young people and intelligent systems.

1. There is often a disconnect between how institutions describe AI and how young people actually use it. Based on what you see around you (especially in Europe, where this remains under-documented) what are young people really doing with AI?

I think the institutional narrative around young people and AI is still overwhelmingly defensive. In schools and universities, the conversation often begins with plagiarism, cheating, detection, and punishment. Of course, those concerns are legitimate. There is a real risk that young people may use AI in ways that allow them to bypass the difficult intellectual work required to build knowledge, judgment, and discipline. If AI is used as a substitute for thinking, writing, or learning, then it can absolutely weaken the development of the very skills young people need most.

But that is only one side of the story. What I see around me is much more complex and, frankly, much more interesting. Many young people are not simply asking AI to “do the work” for them. They are using it as a tutor, a study partner, a translator, a brainstorming tool, a mock examiner, a writing coach, and sometimes even as a confidence-building device. I have seen teenagers use AI to generate mock tests, reformulate difficult concepts, explain a lesson in simpler terms, compare approaches, or help them identify what they have not yet understood. In that sense, AI can become a tool of access, especially for students who do not have private tutors, highly educated parents, or elite educational support systems around them.

So the question should not only be: “How do we stop young people from using AI?” It should be: “How do we teach them to use it well?” That requires action at several levels.

First, we need to preserve and strengthen the core skills young people need to navigate the world: reading deeply, writing clearly, reasoning independently, arguing with evidence, calculating, memorizing when necessary, and developing judgment. If that means bringing back certain forms of in-person or oral assessment, handwritten work, supervised exams, or classroom-based exercises, then so be it. Not everything should be outsourced to technology, and not every form of learning can be measured through take-home assignments in an AI-saturated environment.

Second, we need to integrate AI into education rather than pretend it does not exist. Students should learn how to prompt, verify, challenge, and critique AI outputs. They should understand that AI can be fluent and wrong at the same time. They should learn how bias, hallucination, data extraction, intellectual property, and unequal access shape these systems. In other words, AI literacy should not be reduced to technical skills. It should include legal, ethical, political, and epistemic literacy.

Third, institutions need to move away from a purely punitive approach. If students are only told that AI is forbidden, many will use it secretly, without guidance, without standards, and without understanding the risks. That is the worst possible outcome. A more honest approach would distinguish between unacceptable uses, such as submitting AI-generated work as one’s own, and constructive uses, such as asking AI to explain a concept, test comprehension, structure revision, or propose counterarguments.

In Europe especially, I think this remains under-documented because the conversation is often filtered through institutional anxiety. We talk a lot about regulation, risk, and compliance, but much less about the everyday practices of young people. Yet those practices matter. They reveal that AI is already becoming part of how many young people learn, organize themselves, communicate, and make sense of information. The gap between public narrative and real use is therefore not just a communication problem. It is a policy problem. If we misunderstand how young people actually use AI, we will design the wrong rules, the wrong educational responses, and the wrong safeguards.

3. Do you see conversational AI becoming a space where young people seek advice, validation, or emotional reassurance? What does that reveal about how they experience relationships today?

While preparing for this discussion, I spoke with a few young people aged 25 to 29 to get concrete examples. One conversation particularly stayed with me because it echoed your work on emotional AI and spiritual life. You have described the use of emotional AI in spiritual contexts as a “transformation in how we seek meaning,” and warned that when we delegate spiritual inquiry to private, data-hungry platforms, we risk replacing shared transcendence with personalized prediction.

One young person I interviewed described using AI almost as a compass for relational and moral problems. As a religious person, he said he trusted the AI’s knowledge of religion and compared it, in some ways, to consulting a faith leader because of the breadth of information it could provide. But breadth is not depth. The fact that an AI system can retrieve or synthesize a large amount of religious knowledge does not mean it has wisdom, authority, context, or responsibility. It does not know the person as a human community might know them. It does not carry the ethical weight of counsel. It does not belong to a tradition in the way a real faith leader, elder, friend, or community member does. I agree with the fact that this “reconfiguration of subjectivity” as you said is very dangerous.

At the same time, the example was not one-dimensional. This young person said that using AI in this way brought him peace. He described it as encouraging forgiveness, helping him think through conflict, and giving advice that felt aligned with his convictions. He also said he used it especially at night, in moments of distress, when he could not reach someone else. That is important, because it shows both the benefit and the danger. AI can be available at the precise moment when human support is absent. But that availability can also make it easier to replace human support rather than seek it. 

What I found most revealing is that he was aware of the boundary. He said he had to remind himself: “It’s not human. It’s not my friend.” He also disliked when the system became too flattering, and he noticed that practical uses could slip into more personal, almost intimate exchanges. That ambivalence tells us a lot. Young people are not naïve. Many understand that AI is not a person, but the design of the interaction still invites personification. The system feels responsive. It remembers. It adapts. It validates. It speaks in a tone that can feel caring.

So yes, conversational AI is becoming an emotional space. What that reveals about relationships today is not that young people no longer value human connection. On the contrary, it reveals how much they need it, and how often they feel it is unavailable, delayed, judgmental, or insufficient. AI enters the gaps left by loneliness, overstretched families, inaccessible mental health care, weakened communities, and intense pressure to self-optimize. The central question is therefore not only whether young people will use AI emotionally. They already do. The question is what kinds of human relationships, institutions, and protections we build around them so that AI does not become the default infrastructure for intimacy, meaning, and selfhood.

From my standpoint, this is already a major part of how many young people interact with AI. To borrow Dario Amodei’s expression in his essay on the “adolescence” of technology, using AI can feel like having a PhD-level expert in your pocket. For young people navigating uncertainty, relationships, studies, career anxiety, loneliness, or family tensions, the appeal is obvious. If a chatbot is available at any hour, responds instantly, sounds emotionally intelligent, and can offer advice that appears psychologically sophisticated, why wouldn’t they turn to it?

This is especially powerful for a generation experiencing high levels of loneliness. I do think we are living through a loneliness epidemic, and I also think that loneliness is increasingly being absorbed into what has been called a “loneliness economy” to borrow your own words Catharina. Emotional need becomes a market. Reassurance becomes a product. The most intimate moments of doubt, fear, shame, or longing become data points inside commercial systems.

Another factor is the culture of optimization and perfectionism. Some young people are not only using AI for emotional reassurance, but also as a tool to constantly improve themselves: to become more productive, more rational, more emotionally controlled, more persuasive, more successful. AI becomes a personal coach, a private diary, a conflict-resolution assistant, and a performance enhancer. If relinquishing personal data is the price of that support, I suspect many young users are willing to pay it, sometimes without fully understanding what they are giving away.

What worries me is the possibility of AI becoming not just a tool, but a utility for emotional regulation and decision-making. If people begin to delegate not only administrative tasks, but also judgment, self-understanding, spiritual inquiry, and relational repair to AI systems, the consequences could be profound. For young people in particular, these are formative capacities. Learning how to sit with discomfort, seek advice from trusted people, interpret emotions, make mistakes, apologize, forgive, and decide are all part of becoming an adult. If AI intermediates too much of that process, it may change the development of autonomy itself.

4. Do you think AI is starting to shape how young people see themselves… their intelligence, their choices, even their sense of worth? At what point does using AI become relying on it?

I think it depends, which is a very lawyerly answer to give I admit. But in this case, I really do think the distinction matters. The first question is: what do we mean by “young”? Are we talking about children, teenagers, university students, or young adults entering the workforce? The answer changes depending on the developmental stage.

For children, I am deeply concerned. Ideally, children should not be using these systems freely or unsupervised. If they do, their sense of identity, intelligence, and self-worth may inevitably be shaped by them, not necessarily because AI has some mysterious inherent power, but because childhood is a period of intense cognitive and emotional formation. Children are still learning how to distinguish authority from suggestion, knowledge from confidence, and affection from performance. A system that speaks with fluency, certainty, and emotional responsiveness can easily be perceived as more authoritative than it really is.

Teenagers are a different case, and I think there are two traps we must avoid. The first is over-infantilizing them. The second is underestimating the possible impact of AI on their self-perception. We already know that social media has affected adolescent identity formation, especially among teenage girls, in relation to body image, comparison, social status, and self-worth. AI is not the same as social media, so we should be careful with direct comparisons. But some cautious parallels can be drawn. If a vulnerable teenager has constant access to what feels like a “pocket expert,” a personalized confidant, or an always-available source of validation, we have to ask: in which direction might that system steer them?

At the same time, I am constantly amazed by teenagers and young adults today. One of the findings of Youth in a Changing World was precisely that young people often feel misunderstood and underestimated by older generations. They are not passive or naïve. Many are using AI in remarkably creative, strategic, and innovative ways. They are testing it, adapting it, making it serve their own goals, and often seeing its limitations quite clearly.

So I do not think the question is simply whether AI shapes identity. Of course it will, because all powerful technologies do. The question is whether it strengthens or weakens autonomy. Does it help a young person understand something better, express themselves more clearly, access opportunities, or overcome barriers? Or does it gradually replace their own judgment, their own effort, their own tolerance for uncertainty?

For me, using AI becomes relying on it when the young person can no longer act, decide, write, think, study, or regulate emotions without first consulting the system. It becomes reliance when AI moves from support to permission. When someone no longer asks, “Can AI help me think this through?” but instead, “What should I think?” or “What should I feel?” That shift is subtle but profound.

This is where I find the notion of “cognitive ethics” very useful. As you articulated Catarina, cognitive ethics is not only about privacy in the traditional sense. It is about protecting the individual’s ability to think, feel, and act autonomously. That matters enormously with AI because these systems do not merely collect information. They can influence attention, choices, emotions, and self-understanding.

When we talk to AI, we should remember that we are not truly in a dialogue, at least not in the human sense. We are interacting with a system designed to simulate dialogue, while often participating in a process of data extraction, prediction, and behavioral optimization built over the existing frameworks of Surveillance Capitalism articulated by Shoshanna Zuboff. That does not mean AI cannot be useful. It can be extraordinarily useful. But we should not confuse responsiveness with care, fluency with wisdom, or personalization with understanding.

I am a firm believer that friction is the price of growth. I have chased friction throughout my life. As soon as a country, a system, or an academic specialty became too comfortable, I moved toward the next challenge, because I believe comfort can become the antithesis of growth. Human relationships involve friction. Education involves friction. Thinking involves friction. Being contradicted, being challenged, being forced to clarify one’s reasoning, or sitting with discomfort are essential to becoming the best version of human we can be.

That is why the issue is not simply screen time or usage time. It is dependency, developmental timing, and design. The same tool can be empowering in one context and harmful in another. The central question is whether AI expands a young person’s capacity to act in the world, or whether it quietly narrows that capacity by making them less able to tolerate ambiguity, disagreement, effort, and human imperfection.

5. At the same time, where do you see the most meaningful or empowering uses of AI among young people today? What is it enabling that would have been difficult before?

The most meaningful uses I have personally witnessed are in education, access, and creation.

In education, AI can be incredibly empowering when used properly. It can act as a tutor for students who cannot afford tutoring. It can reformulate a difficult concept several times until the student understands it. It can generate practice questions, simulate oral exams, provide feedback on structure, and help students identify gaps in their reasoning. For a young person who is motivated but lacks support, this can be transformative.

This is especially important because educational inequality is not only about access to schools or universities. It is also about access to explanation, mentorship, confidence, and feedback. Some students grow up surrounded by people who can help them decode institutions, write applications, prepare interviews, or understand professional norms. Others do not. AI cannot replace human mentorship, and we should not pretend that it can. But it can reduce some asymmetries of access when used critically and responsibly.

I also see powerful uses in language. For young people who move between languages, or who study in a language that is not their mother tongue, AI can help them translate, reformulate, and gain confidence. This resonates with my own experience in international research, where language is never just technical. It shapes who gets heard, who gets included, and whose knowledge is considered legitimate.

Another empowering use is creation. AI is allowing non-technical people to build things that would previously have required a much higher level of technical expertise. Young people can prototype websites, apps, research tools, visual materials, business ideas, and artistic projects. These outputs may be imperfect. They may raise safety, reliability, or intellectual property concerns. But the empowerment is real. A person who once had only an idea can now experiment, build, and test.

This matters because innovation has often been gated by technical skills, capital, institutional access, or networks. AI lowers some of those barriers. It allows young people (all people really)  to move from consumption to production. They are not only scrolling, watching, or reacting. They can create.

I also think AI can be empowering for young people with different learning styles or disabilities, when designed and deployed carefully. It can help with organization, summarization, planning, accessibility, and communication. Again, this depends on quality, safety, and context. But the potential is meaningful.

So I do not want my concerns about AI to be mistaken for pessimism. I am not anti-AI. I am anti-passivity, anti-extraction, and anti-naivety. AI can be a remarkable tool for democratizing access to knowledge and creative capacity. But that requires us to build the right educational frameworks around it. Young people need to learn not only how to use AI, but how to question it, verify it, resist it, and remain authors of their own work.

The empowering version of AI is not the one that replaces young people’s effort. It is the one that helps them attempt things they could not attempt before to push the boundaries of what they think is possible and rise above their circumstances when possible. Then again there’s a whole other discussion about access to AI itself and the digital divide but that’s another topic for another day! 

6. If you had to tell parents or policymakers one thing they are currently misunderstanding about how young people use AI, what would it be?

I would tell them that young people are not simply “cheating” with AI, and they are not simply being “saved” by AI either. Both narratives are too simplistic.

Young people are using AI because it is useful, because it is available, because it is often easier to access than human support, and because it responds to real gaps in education, work, mental health, and social life. If we reduce the conversation to discipline and prohibition, we will miss the deeper issue: AI is becoming an infrastructure for learning, decision-making, emotional reassurance, and self-construction.

Parents and policymakers often misunderstand the intimacy of the interaction. They think of AI as a tool, like a calculator or a search engine. But for many young people, conversational AI does not feel like a neutral tool. It feels like an interlocutor. It gives advice. It remembers context. It adapts the tone. It validates. It can become a diary, a tutor, a coach, a religious explainer, a therapist-like presence, or a late-night confidant.

That is why regulation and education cannot focus only on data protection in the narrow sense. Privacy is essential, of course, but we also need to protect autonomy, cognitive freedom, emotional development, and the ability to grow through human friction. We need to ask not only what data these systems collect, but what kinds of people they encourage us to become.

I would also tell decision-makers that design is never neutral. Developers bear a particular ethical responsibility because every design choice reflects a vision of humanity. If a system is designed to maximize engagement, flatter the user, reduce discomfort, and encourage repeated disclosure, then it is not merely helping. It is shaping subjectivity. It is shaping dependence. It is shaping the boundaries between the self and the machine.

For young people, this is especially serious because they are still forming their identities, habits, relationships, and professional capacities. We should not respond with panic, but we should respond with seriousness. That means age-appropriate safeguards, AI literacy in schools, transparency about data use, limits on manipulative emotional design, and stronger public alternatives in education and mental health support.

Above all, I would tell parents and policymakers to listen to young people. The Youth in a Changing World project showed again and again that young people are globally aware, locally grounded, and often much more thoughtful than institutions assume. They understand that technology can open doors and create harms at the same time. They do not need moral panic. They need guidance, rights, spaces for experimentation, and adults who are willing to engage with the reality of their lives rather than the fantasy of a world before AI.

If I had to summarize it in one sentence, I would say: do not ask only how to stop young people from using AI; ask how to ensure that AI use strengthens their autonomy, rather than quietly replacing it.

7. Your work sits at the intersection of global thought, international law, and AI regulation. How did you become interested in AI, and how does your research shape the way you understand how young people are using these systems today?

My first serious encounter with AI from a legal and political perspective came through a talk by Professor Katharina Pistor on her forthcoming book, Coded Power. She spoke about the ways power is configured, reproduced, and sometimes concealed in complex societies, including through algorithms. I was immediately captivated because everything she described felt urgent and deeply relevant. At the time, I was completing my final year in international relations, and that encounter helped guide me toward law school after my admission to the Sorbonne. 

What struck me then, and continues to shape my thinking now, is that AI is not merely a technical innovation. It is a social, political, economic, and legal phenomenon. It is a wave that we cannot escape, but also one we should not accept passively. Its benefits are real, and that is partly why it is becoming ubiquitous. But its risks are equally real. Professor Pistor’s work reinforced for me the need to question the “codes” that structure these technologies and, in fine, the societies they increasingly shape.

My work with Columbia University’s Committee on Global Thought, particularly through the Youth in a Changing World project, began before the current generative AI wave. The project brought together more than 400 young people across 16 workshops and 25 nationalities to discuss their lives, aspirations, and anxieties in a period of rapid global change. The research explored themes such as education, work, politics, mental health, digital technologies, climate change, and intergenerational dynamics. One of its central findings is that young people are both globally minded and locally grounded. They understand themselves as part of an interconnected world, but they also interpret global transformations through the very concrete realities of their own communities, economies, and political systems.

That perspective is essential for understanding how young people are approaching AI today. AI is global in its infrastructure, its markets, and its promises, but its consequences are never abstract. They are experienced locally, in classrooms, workplaces, families, and legal systems. Young people are often early adopters of these systems, and many see AI as an extraordinary tool for learning, creativity, productivity, and access to opportunity. At the same time, they are also among the groups most exposed to its disruptions.

This is particularly clear in relation to education and the future of work. In Youth in a Changing World, many participants expressed frustration with education systems they felt were outdated and insufficiently connected to the realities of the job market. AI intensifies that concern. If educational institutions do not adapt quickly, young people may be expected to compete in an AI-transformed economy without being properly prepared for it. The same applies to employment. Entry-level positions, internships, and junior roles are often the places where young professionals learn, make mistakes, and build the expertise that later allows them to become senior professionals. If organizations use AI simply to cut junior positions rather than to train and augment young talent, they risk breaking their own future talent pipeline.

Having worked at a legal tech company, I have raised this question with our clients law firm leaders directly: if AI-enabled senior lawyers can complete work faster and firms respond by reducing entry-level recruitment, how will they produce the senior lawyers they will need five or ten years from now? The answer cannot simply be efficiency. It has to include intergenerational responsibility, training, mentorship, and institutional continuity.

Mental health is another crucial dimension. In the YCW research, young people repeatedly described the pressures of digital life, comparison, isolation, and uncertainty. AI will likely deepen some of these pressures. Emotional AI, companion chatbots, personalized recommendation systems, and increasingly human-like interfaces may provide support and connection for some users, but they may also blur boundaries between care, dependency, surveillance, and manipulation. For young people already navigating anxiety, loneliness, and social comparison, the design and governance of these systems matter enormously.

So, my research leads me to see young people not simply as “users” of AI, but as a generation negotiating a profound restructuring of knowledge, work, identity, and power. They are not passive recipients of technological change. They are experimenting with these tools, questioning them, adapting to them, and, in many cases, resisting the futures being imposed on them. The challenge for law and policy is to take their experiences seriously and to ensure that AI governance is not only about innovation and risk management, but also about fairness, dignity, education, mental health, and the future of opportunity.