Anglais

In a recent supervision session, a trainee shared an idea for a simple tool to help a patient track mood between sessions. Not long ago, this idea would likely have stayed theoretical unless a developer was involved. Today, with OpenAI just releasing Codex Sites, a written idea can be turned into a working app that can be shared through a simple link. This changes not only what we can build, but how we think about applying ideas in clinical care. What is new is not just faster coding, but a simpler process overall. In the past, building a tool required planning, design, programming, testing, and hosting. Each step required time and often different skills. Now, much of this can happen through a guided conversation with a system. This reduces the distance between having an idea and seeing it in action. From a clinical perspective, this can support how we think and reason. The concept of the “extended mind” suggests that tools can help us think more effectively (Clark & Chalmers, 1998). When clinicians can quickly turn ideas into small working tools, they can test, reflect, and refine their thinking. This may encourage a more active and flexible approach to problem-solving in practice. This shift has practical value across different areas. A therapist might create a simple app for mood tracking, coping reminders, or session feedback. A researcher might design a tool to collect data in a more tailored way. Instead of relying only on standard tools, clinicians can begin to shape tools that better fit their patients’ needs and contexts. However, ease of creation does not guarantee quality. A tool that works technically is not always clinically appropriate. Evidence-based practice still requires theory, research, and careful judgment (Sackett et al., 1996). Without this foundation, there is a risk of creating tools that are engaging but not effective, or even misleading. Looking at fields like design science can offer useful guidance. New tools are often developed in small steps, tested, and improved over time (Hevner et al., 2004). Clinicians can adopt a similar mindset, while maintaining attention to safety and validity. Iteration is valuable, but it must be guided by clinical knowledge and patient well-being. These developments may also change how clinicians see their role. Some may begin to act not only as practitioners, but also as creators of simple digital tools. This can feel empowering, but also unfamiliar. Many clinicians do not have formal training in technology, so support and education will be important for responsible use. Ethical considerations are central in this process. When clinicians create or use digital tools, they are responsible for how these tools affect patients. Transparency about how a tool works, how data is handled, and what its limits are is essential. Patients should clearly understand what they are using and how it may influence their care. When AI is involved, additional caution is needed. These systems can produce errors, reflect bias, or generate outputs that appear reliable but are not well supported. Clinicians must remain critical and thoughtful in their use. While the gap between idea and application is shrinking, clinical responsibility and careful judgment remain unchanged.

Anthropic, the company behind Claude, widely considered one of the most capable AI systems available today, did not build seven completely different models. What they built is simpler than it first appears. It is one system, organized as a ladder. The names can sound confusing at first, but in practice there are only five levels to understand, moving from fast and efficient to increasingly powerful forms of reasoning. Once you understand this ladder, the system becomes much easier to use. A therapist recently said, “I keep hearing different names, Haiku, Sonnet, Opus, and I don’t know which one I should use.” This is a very common experience. The difficulty is not the technology itself, but how it is presented. When there are too many names, it starts to feel complicated very quickly. A more helpful way to understand Claude is this: it is not many separate AI systems. It is one system organized as a ladder. Each name represents a different level of ability, moving from fast and simple to more powerful and complex. Once you see it this way, the confusion becomes much easier to manage. At the lowest level is Haiku. This is the fastest and most affordable option. It works well for simple, repetitive tasks that do not require deep thinking. For example, a therapist might use Haiku to summarize session notes, rewrite a paragraph, or organize brief information. It is helpful for saving time, but it is not designed for complex clinical reasoning. The next level is Sonnet, and this is the model most clinicians will find useful in everyday work. It offers a good balance between quality and cost. Sonnet can support tasks like writing case notes, developing treatment plans, or creating psychoeducational material. It is reliable and clear, without being too resource-intensive. For many therapists, this becomes the default choice. Above that is Opus, which is designed for more demanding thinking. This is the model to use when the work becomes more complex, such as exploring case formulations, comparing diagnostic possibilities, or integrating research into practice. It can handle more depth, but it also requires more resources, so it is usually best used when needed rather than all the time. At the top of the ladder are Fable and Mythos. These are not two different systems, but the same model with different settings. Fable includes safety guardrails, meaning it is more cautious when responding to sensitive or high-risk topics. Mythos has fewer restrictions and allows more open responses, but it is typically limited to expert use. For most clinical settings, Fable is the more appropriate and responsible choice. However, a recent development adds a key limitation: the US government has issued an export control directive suspending access to Fable 5 and Mythos 5 for foreign nationals, both inside and outside the United States. As a result, these systems may be abruptly unavailable to many users in practice. Thinking in terms of a ladder can help guide decisions in practice. Instead of asking, “Which model is best?”, it is more useful to ask, “How much support do I need for this task?” Simple tasks can stay at the lower levels, while more complex clinical questions may require moving higher. This approach is similar to how therapists already adjust interventions based on client needs. It is also important to remember that these tools support thinking, but do not replace clinical judgment. AI can help organize ideas, suggest possibilities, or clarify language, but it does not fully understand the client or the therapeutic relationship. The clinician remains responsible for interpreting and deciding what is appropriate. There are also ethical considerations to keep in mind. When using AI in clinical or research settings, it is important to be transparent about how it is used and to ensure that sensitive information is handled carefully. Clinicians should also be aware that AI systems can reflect biases or make errors, even when responses sound confident. Careful review and critical thinking are always necessary. As these tools become more common, they may begin to shape how clinicians write, think, and communicate. This creates new opportunities, but also new responsibilities. It will be important to continue reflecting on how AI fits into clinical practice, rather than using it automatically or without question. In the end, the goal is not to master every model or feature. It is simply to understand the basic structure: one ladder, with different levels of support. With this perspective, AI becomes less overwhelming and more practical. It can then serve as a helpful extension of clinical work, while the therapist remains at the center of decision-making and care.

Dans la pratique clinique quotidienne, de nombreux thérapeutes commencent à utiliser des outils d'IA pour aider à rédiger des notes, à organiser des informations ou à réfléchir aux diagnostics. Ces outils peuvent être utiles, mais ils soulèvent également une question importante: comprenons-nous vraiment comment ils fonctionnent? Le lancement d'Apertus, un modèle multilingue d'IA open source développé en Suisse, nous invite à réfléchir non seulement à ce que l'IA peut faire, mais aussi à ce que nous pouvons lui faire confiance. Une idée clé derrière Apertus est la souveraineté numérique, ce qui signifie avoir le contrôle sur l'endroit où les données sont stockées et comment elles sont utilisées. Pour les thérapeutes, c'est particulièrement important parce que nous travaillons avec des informations sensibles sur les patients. Lorsque les systèmes d'IA appartiennent à de grandes entreprises à l'étranger, il n'est pas toujours clair comment les données sont traitées. Apertus vise à offrir un contrôle plus local, en s'harmonisant mieux avec les responsabilités éthiques et juridiques dans les soins de santé. De nombreux outils d'IA existants, comme ChatGPT, sont puissants mais pas entièrement transparents. Nous pouvons voir leurs sorties, mais pas facilement comprendre comment elles sont produites. Pour les cliniciens formés à raisonner soigneusement, cela peut se sentir mal à l'aise. Apertus offre une approche plus ouverte, permettant aux experts d'examiner et de comprendre le fonctionnement du système, ce qui peut favoriser une utilisation plus éclairée. Une autre caractéristique importante d'Apertus est sa capacité multilingue. En thérapie, le langage joue un rôle clé dans la façon dont les patients s'expriment. Un système qui fonctionne dans toutes les langues peut aider à réduire les malentendus et soutenir des soins plus inclusifs. Toutefois, le contexte culturel demeure complexe et aucun système d'IA ne peut le saisir pleinement. Parce qu'Apertus est open source, r

Are we ever going to stop the development of AI? As these systems become increasingly intelligent, a deeper question emerges: when do we decide that enough is enough for humanity? In May 2026, Pope Leo XIV contributed to this global debate by warning that artificial intelligence must not take moral decision-making out of human hands. While his message speaks broadly, it resonates strongly within clinical practice, where efficiency, innovation, and ethical responsibility are already tightly intertwined. In everyday care, AI often appears in subtle ways. It summarizes patient histories, drafts reports, or suggests diagnoses. These tools can ease administrative burden, but their influence is not neutral. Over time, they begin to shape how clinicians organize information and approach decisions. This raises a critical question: where does clinical judgment truly reside? Clinical reasoning has traditionally been a reflective process grounded in human experience and patient interaction. As AI provides ready-made interpretations, there is a growing risk that parts of this reasoning become less deliberate, even if unintentionally. Responsibility, however, remains fully human. Clinicians are still accountable for evaluating and deciding whether to trust AI-generated outputs. This requires active engagement, asking not only “what does this suggest?” but also “why, and should I rely on it?” Beyond healthcare, some experts and voices within the AI industry have already called for slowing development, emphasizing the need for society to adapt. The debate is no longer just about capability, but about the pace of integration. Transparency adds another layer of concern. Many AI systems cannot clearly explain how they reach their conclusions. In clinical care, where decisions must be understandable, this creates tension, especially as patients seek meaning, not just outcomes. Bias and inequality further complicate the picture. AI systems reflect the data they are built on, which can carry social and cultural biases. At the same time, access to advanced tools remains uneven, raising questions about fairness in care quality. At its core, the rapid expansion of AI challenges the relational nature of clinical work. Therapy is not just about information, it is about presence, attunement, and human connection. Increasing reliance on AI risks filtering patient experiences through predefined systems rather than fully exploring them. So the question is not whether AI should continue to develop, but how far it should go without clearer ethical limits. Stopping it is unrealistic, but moving forward without reflection carries real consequences. For clinicians, this means maintaining an active, critical stance, using AI as support, not substitute, and staying transparent with patients. More broadly, it calls for a collective effort to ensure that innovation does not outpace responsibility. Ultimately, this is not just a technological shift, but a human one. The challenge is not only to understand what AI can do, but to decide, carefully and consciously, where we choose to draw the line.

Joyeux cerveau La formation a été construite pour passer au-delà de la théorie et à une véritable transformation pratique. Au cours de la dernière année, mon travail s'est développé au-delà des frontières, des formats et des communautés, guidé par un objectif clair : aider les thérapeutes à utiliser l'IA de manière à faire vraiment la différence. Ce que je fais est simple dans l'idée, mais puissant dans la pratique. Je forme les thérapeutes à utiliser efficacement l'IA en thérapie, en mettant l'accent sur des applications pratiques et réelles qui améliorent l'efficacité, la créativité et les résultats cliniques. Mes programmes de formation sont structurés comme un voyage. Les thérapeutes commencent par construire une base solide, comprendre l'IA générative et des outils comme ChatGPT, puis progressent vers la création de matériaux thérapeutiques personnalisés, et finissent par passer à des études de cas avancées et à la planification complète de la thérapie soutenue par l'IA. Chaque étape est conçue pour renforcer la confiance, l'autonomie et l'impact clinique. Tout au long de ce voyage, nous explorons également l'ingénierie rapide, la productivité et l'utilisation de l'IA comme outil de soutien pour les patients. Les programmes sont passés d'interactions de base à des matériaux structurés, des techniques et des applications spécifiques à chaque domaine. Des solutions ciblées sont également conçues pour les professionnels qui veulent améliorer l'efficacité et rationaliser leur flux de travail. Au-delà des programmes structurés, je propose des séances de consultation individuelles, offrant des conseils personnalisés adaptés aux défis de chaque thérapeute. Mon approche reste fondée sur de véritables besoins thérapeutiques, de la communication et du langage à des troubles complexes comme l'aphasie, la dysarthrie et la dysphagie. Apprendre ne reste pas dans un seul endroit. Mes formations en personne ont atteint

Joyeux cerveau La formation s'appuie sur une vision claire et ambitieuse : repenser l'évolution de la thérapie dans un monde qui évolue plus rapidement que jamais. Je l'ai fondé en tant que orthophoniste et neuropsychologue avec une solide expertise en IA générative, réunissant la profondeur clinique et l'innovation technologique d'une manière qui se sent à la fois fondée et tournée vers l'avenir. Ce qui a commencé par une idée est devenu quelque chose de beaucoup plus tangible. C'est maintenant un espace où les thérapeutes viennent explorer, questionner et remodeler leur pratique. Avec mon expérience dans le travail clinique, la recherche et l'IA, l'objectif n'a jamais été de simplement introduire de nouveaux outils, mais d'aider les thérapeutes à les utiliser avec clarté, objectif et jugement clinique. Au cœur de ce projet, Happy Brain Training est animé par une mission simple mais puissante : libérer le potentiel des thérapeutes et de leurs patients grâce à l'utilisation réfléchie et responsable de l'IA génératrice. Cela signifie équiper les thérapeutes d'outils qui soutiennent véritablement leur travail, les aider à créer des matériaux plus personnalisés et engageants, et rendre l'IA accessible plutôt qu'écrasante. C'est exactement pourquoi j'ai créé le bulletin Happy Brain Training : tenir les thérapeutes informés, inspirés et connectés au paysage évolutif de la thérapie et de la technologie. Publié chaque semaine, il fournit des mises à jour sur les nouveaux modèles, les nouvelles possibilités et les principaux développements sur le terrain. Il aborde également des sujets essentiels tels que la sécurité des données, la protection de la vie privée, les règlements, les considérations éthiques et même l'impact écologique de l'IA. Dans un domaine qui évolue rapidement, rester informé n'est plus

A therapist looking at an image often focuses on meaning, what it shows, how it feels, and how it connects to a client’s story. Until recently, we could also assume that most images reflected something real. Today, with the rise of AI-generated images, that assumption is less certain. New tools, such as OpenAI’s image verification system, are beginning to help us check where images come from and whether they were created by AI. This tool allows users to upload an image and see if it contains hidden signals linked to AI systems, such as metadata or digital watermarks. These signals can suggest that an image was generated using tools like ChatGPT or related APIs. For clinicians, this is less about technology itself and more about developing a habit of asking: “Can I trust the origin of this image?” In practical terms, using the tool is simple and does not require technical knowledge. You begin by going to the verification webpage and uploading the image you want to check. It is helpful to crop the image so that only the main picture is included, especially if it comes from a screenshot. After uploading, the tool analyzes the file and looks for known signals linked to AI generation. It will then show whether these signals are present. If they are, the image was likely created using AI tools. If not, the result is less certain, and further caution is still needed. This process takes only a few moments and can easily become part of routine checking when working with unfamiliar images. In everyday clinical work, images are used in many ways, psychoeducation, assessment, and even therapeutic exercises. When we assume an image is real, we may respond to it differently than if we knew it was created by AI. This makes it important to pause and reflect, especially when an image plays a role in clinical understanding or emotional processing. From a thinking perspective, this shift asks us to slow down. We often rely on quick impressions when we see an image, especially if it looks familiar or realistic. However, AI-generated images can look highly convincing. Verification tools can support us in moving from quick assumptions to more careful, reflective thinking. There are also learning implications. Students and early-career therapists often use visual materials to support memory and understanding. If an image is later found to be artificial, it may create confusion or reduce confidence. Knowing that verification tools exist can help build a more balanced approach, one that combines curiosity with healthy doubt. In research settings, the issue becomes even more important. Fields that depend on images, such as medical or rehabilitation sciences, require accurate and trustworthy data. If AI-generated images are used without clear labeling, this can affect the quality of research. Verification tools can support better data practices, but they are only one part of the process. It is also important to understand the limits of these tools. Not all AI-generated images contain detectable signals, and not all real images are easy to verify. A tool may suggest that an image is AI-generated, but it cannot fully explain how or why it was created. This means we should use these tools as support, not as final proof. In clinical practice, ethical questions naturally arise. If a therapist uses an AI-generated image, should they tell the client? In most cases, transparency helps maintain trust. Even if the image is helpful, its origin still matters. Being open about this can strengthen the therapeutic relationship rather than weaken it. Responsibility remains with the clinician or researcher. Tools can assist, but they do not replace professional judgment. This includes thinking carefully about how images are used, checking their sources when needed, and being honest about their origins. It also means being aware that AI systems can carry biases or produce misleading content. As these technologies become more common, clinicians and researchers will need to adapt without feeling overwhelmed. The goal is not to become an AI expert, but to stay thoughtful and attentive. By combining simple verification tools with good clinical judgment, we can continue to use images in ways that are both effective and responsible. Looking ahead, these tools may become a normal part of practice, much like checking references or reviewing data sources. They remind us that in a digital world, seeing is not always the same as knowing. What remains essential is our ability to reflect, question, and make careful decisions in the service of those we work with. To explore the verification tool, visit: https://openai.com/research/verify/

A recent experimental study examined how tools like ChatGPT influence thinking during writing A recent experimental study by Kosmyna, N. et al. (2025), conducted at the MIT Media Lab, explores how tools like ChatGPT influence human thinking during writing tasks. The study, titled “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task,” examined how different levels of technological assistance shape cognitive engagement. The researchers recruited participants and divided them into three groups: one using an AI assistant (LLM group), one using a search engine, and one relying only on their own thinking (Brain-only group). Across three sessions, each group used the same method. In a fourth session, roles were reversed for some participants: those who had used AI were asked to work without it (LLM-to-Brain), while those who had worked independently were introduced to AI (Brain-to-LLM). The study combined EEG brain recordings, language analysis, essay scoring, and participant interviews to understand not just performance, but underlying cognitive processes. The results showed clear differences in how participants engaged cognitively. Brain activity, measured through neural connectivity, was strongest in the Brain-only group, moderate in the Search Engine group, and weakest in the LLM group. This suggests that as external support increased, internal cognitive engagement decreased. In parallel, language analysis revealed that essays produced with AI were more similar to each other, showing less variation in vocabulary and structure, while independently written essays were more diverse and distinct. Participants’ experiences also reflected these differences. Those in the LLM group reported a lower sense of ownership over their essays and had more difficulty recalling or quoting what they had just written. In contrast, the Brain-only group showed strong memory recall and a clear sense that the work belonged to them. Even when AI-assisted essays scored well, they often required minimal editing and remained close to default AI-generated responses, indicating lower levels of active processing. The fourth session provided some of the most important insights. Participants who moved from Brain-only to AI use (Brain-to-LLM) showed increased brain connectivity across multiple frequency bands, suggesting active integration of AI support with prior knowledge. They also performed well in terms of memory and structure. However, those who moved from AI use to independent writing (LLM-to-Brain) showed reduced neural engagement and did not return to the same level of cognitive activity as the original Brain-only group. Their writing also showed traces of AI-influenced vocabulary and structure, indicating a lingering effect of prior AI use. From a clinical perspective, these findings are highly relevant. Clinical reasoning depends on active engagement, organizing information, making connections, and reflecting on decisions. Writing is one of the main ways clinicians develop and refine this reasoning. If AI reduces the need for this effort, especially early in training, it may lead to what can be described as cognitive debt: a gradual weakening of the thinking processes that support clinical judgment. At the same time, the study suggests that AI can be beneficial when used after independent thinking has been established. The Brain-to-LLM group demonstrated that prior effortful engagement may allow clinicians or students to use AI in a more integrated and reflective way. This aligns with educational and clinical models where support tools are most effective when they build on an existing foundation rather than replace it. These findings also echo everyday clinical practice. Therapists often emphasize the importance of active participation and reflection in patients. Similarly, clinicians themselves rely on repeated, effortful thinking to build expertise. If AI tools begin to replace rather than support this process, there may be subtle but meaningful changes in how clinicians think, remember, and make decisions. The ethical implications are important. Clinicians remain responsible for their reasoning and documentation, even when AI is involved. The reported decrease in perceived ownership raises concerns about reduced critical engagement. If a clinician feels less connected to what they have written, they may be less likely to question it. There are also broader concerns about bias and accuracy, as AI-generated content may not always align with individual patient contexts or cultural considerations. For researchers and students, similar risks apply. High-quality writing is not only about clarity but about understanding. If AI assists in producing text without deep engagement, there is a risk of creating work that appears strong but lacks true comprehension. Maintaining intellectual integrity requires active involvement in the thinking process, not just the final output. Overall, this study offers an important early perspective on how AI tools like ChatGPT may shape cognition over time. For clinicians and therapists, it highlights the need for a balanced approach, one that uses AI as a support while preserving the effortful thinking that underpins clinical expertise. The goal is not to avoid these tools, but to use them in ways that strengthen, rather than replace, the cognitive processes at the heart of learning and practice.

Combien de fois avons-nous souhaité pouvoir tout tenir en même temps ? Tous les rapports, toutes les notes, tous les petits détails qui semblent importants mais qui sont répartis entre les dossiers et les moments. Et si le système que nous utilisons pouvait « se souvenir » plus, pas seulement quelques pages, mais des histoires entières ? Google de Gemini se déplace dans cette direction. Avec la capacité de traiter jusqu'à 2 millions de jetons, il peut prendre dans ce qui serait à peu près 1,5 million de mots. Cela pourrait inclure des notes thérapeutiques, des évaluations, des vidéos, des entrevues et de la recherche, le tout à la fois. Pour les thérapeutes, cela touche quelque chose de très familier : notre travail est rarement basé sur une seule information, mais sur le nombre de pièces réunies. Pensez à un enfant suivi pendant plusieurs années. Différents professionnels, différents rapports, différentes perspectives. Nous passons souvent d'un document à l'autre, en essayant de les relier à un tableau significatif. Un système comme celui-ci pourrait aider à mettre tout en évidence dans un seul endroit et à mettre en évidence des modèles ou des lacunes. Mais même alors, ce n'est pas "comprendre" l'enfant. C'est organiser l'information, pas sentir le sens. Parce que dans le travail clinique, plus de données n'égalent pas une compréhension plus profonde. Nous écoutons, nous observons, nous sentons l'atmosphère dans la pièce. Nous remarquons ce qui est dit, mais aussi ce qui n'est pas dit. L'IA peut se rassembler et se structurer, mais elle ne connaît pas la relation. Il ne ressent pas l'hésitation, la résistance, ou le changement de ton. Utilisé avec soin, ce type d'outil peut encore nous soutenir. Nous pourrions lui demander de comparer des rapports, de résumer des objectifs répétés ou de signaler des incohérences. Elle peut nous aider à nous préparer et à gagner du temps.

A therapist sits with a young adult during a counseling session. The client suddenly becomes quiet, then says they feel overwhelmed but cannot explain why. There is a long pause. The therapist notices the tension, gently acknowledges it, and gives the client space to speak without pressure. Moments like this are at the heart of clinical training. They remind us that learning to be a clinician is not only about knowledge, but also about how we respond to real human emotions. AI is already part of our clinical learning, but it is integrated in different ways across systems. In many universities and training programs, it is now included in both formal education and continuous professional development. Today, there are many tools that clinicians and students can use in their ongoing learning, from conversation simulators to note-writing assistants and case review platforms. This is not something coming in the future, it is already shaping how we learn. These tools can be helpful. They allow students to practice conversations, receive quick feedback, and repeat exercises as often as needed. For learners who benefit from repetition or flexible pacing, this can make education more accessible and personalized. However, learning with AI is not the same as learning with real people. AI can simulate a conversation, but it does not truly feel emotions or respond with genuine presence. In real clinical situations, patients may be silent, overwhelmed, resistant, or unclear. These moments require more than technique. They require attunement. The same applies to education itself. A trainer or supervisor does more than provide information. They sense when a student is struggling, disengaged, or not fully understanding. They adjust their teaching in real time, sometimes slowing down, sometimes challenging, sometimes simply being present. AI cannot sense when someone is emotionally unavailable, anxious, or disconnected. This human sensitivity is a key element of clinical training. For example, a student practicing a difficult conversation with AI might learn useful phrasing. But in a live supervision session, a trainer might notice hesitation in the student’s voice, or discomfort in how they approach a sensitive topic. The trainer can pause, ask reflective questions, and help the student explore what is happening internally. This kind of learning goes beyond performance, it builds clinical awareness. There is also a concern that students may rely too much on AI. If they consistently turn to it for answers, they may not fully develop their own clinical thinking. Good clinicians do more than respond, they question, reflect, and tolerate uncertainty. AI can generate quick answers, but it does not replace the deeper thinking that develops through experience and guided reflection. At the same time, AI can still support learning when used thoughtfully. Students can practice before real encounters, compare AI-generated responses with theoretical models, and discuss them with supervisors. In this way, AI becomes a tool that supports critical thinking rather than replacing it. It is also important to recognize that AI is not neutral. It is trained on large datasets that may include biases. As a result, it may not always reflect diverse cultural, linguistic, or personal experiences accurately. In clinical work, where understanding each individual context is essential, this is a significant limitation. Responsibility remains with the clinician. Even when AI is used, students and professionals must be able to justify their decisions and understand the reasoning behind them. AI should assist learning, not replace professional judgment. For educators, this moment is reshaping how we design and deliver training. Curricula are being adapted, not only to include AI tools, but also to teach how to use them critically and ethically. This includes setting boundaries, protecting confidentiality, and helping students understand both the strengths and the limits of these technologies. In the end, AI may support clinical education, but it cannot replace the human connection at its core. In both therapy and training, what matters most is presence, empathy, and the ability to sense what is happening in another person. These are not technical skills alone, they are relational capacities that develop through real human experience. As education continues to evolve, the challenge is not whether to use AI, but how to integrate it without losing what makes clinical work deeply human.