In everyday clinical practice, many therapists are beginning to use AI tools to help with writing notes, organizing information, or reflecting on diagnoses. These tools can be useful, but they also raise an important question: do we really understand how they work? The launch of Apertus, a multilingual open-source AI model developed in Switzerland, invites us to think not only about what AI can do, but also about how much we can trust it.
One key idea behind Apertus is digital sovereignty, which means having control over where data is stored and how it is used. For therapists, this is especially important because we work with sensitive patient information. When AI systems are owned by large companies abroad, it is not always clear how data is handled. Apertus aims to offer more local control, aligning better with ethical and legal responsibilities in healthcare.
Many existing AI tools, like ChatGPT, are powerful but not fully transparent. We can see their outputs, but not easily understand how they are produced. For clinicians trained in careful reasoning, this can feel uncomfortable. Apertus offers a more open approach, allowing experts to examine and understand how the system works, which may support more informed use.
Another important feature of Apertus is its multilingual ability. In therapy, language plays a key role in how patients express themselves. A system that works across languages may help reduce misunderstandings and support more inclusive care. However, cultural context remains complex, and no AI system can fully capture it.
Because Apertus is open source, researchers and developers can study, adapt, and improve it. This supports transparency and collaboration, but it also means that different versions of the system may exist, which can make standardization more difficult in practice.
Apertus is also designed as a public good, meaning it is freely accessible. This could make advanced AI tools more available to therapists and institutions with limited resources. However, maintaining such a system requires ongoing support, which remains a challenge.
The institutions behind Apertus, EPFL and ETH Zurich, bring strong scientific credibility. Still, in clinical settings, trust develops over time. Therapists need to see how a tool performs in real situations before fully relying on it.
Ethically, the use of AI in therapy requires careful attention. Even with transparent systems, clinicians remain responsible for their decisions. AI should support, not replace, clinical judgment. It is important to question outputs, understand limitations, and remain aware of possible biases.
Apertus represents more than a new tool, it reflects a different approach to AI. It encourages clinicians to engage actively with technology rather than use it passively. As AI continues to evolve, the challenge will be to keep clinical responsibility, critical thinking, and patient trust at the center of practice.