Switzerland and AI: Building responsible, open, and sovereign innovation
In this interview, Professor Martin Jaggi of EPFL presents Switzerland’s emerging role in the field through the Swiss AI Initiative and its flagship model Apertus, highlighting a vision built on openness, multilingual capability, and technological sovereignty. He also discusses the benefits and limits of AI, as well as the need for responsible development and human oversight.
In this interview, Professor Martin Jaggi of EPFL presents Switzerland’s emerging role in the field through the Swiss AI Initiative and its flagship model Apertus, highlighting a vision built on openness, multilingual capability, and technological sovereignty. He also discusses the benefits and limits of AI, as well as the need for responsible development and human oversight.
AI has rapidly entered everyday life for millions of people. Do you believe we can already describe this as a technological revolution?
It is moving fast, and in my opinion can be counted as a revolution, in the sense that it has had significant and far-reaching impacts that fundamentally transform societal structures and behaviors.
Some executives view AI as a kind of “magic tool” that can solve every problem. What would you say to them about their expectations of what AI can realistically deliver?
The tricky bit is the “jagged” profile of the capabilities of AI. While it reaches superhuman level in several areas (such as mathematics and software engineering), it can fall short in many unexpected ways, sometimes on problems that are easy for humans. It is important to carefully measure capabilities and limitations, and their progression over time as models evolve, and to establish practical ways for human oversight and collaboration.
One of the most persistent questions is whether AI will replace human workers and contribute to rising unemployment. What is your perspective on AI’s long-term impact on employment?
The long-term impact on employment is likely to be complex and varied, with job categories emerging and existing ones evolving. While some work will be automated, new jobs will emerge in AI development, deployment, and oversight. It is crucial to invest in education and retraining programs to ensure workers can adapt.
Ethics in AI is another topic that continues to generate debate. Is it realistic to build truly ethical AI systems? And what practical steps can companies take if they want to develop AI responsibly?
Building truly ethical AI systems involves acknowledging the inherent biases in data and algorithms, implementing transparency measures, and establishing ethical guidelines and regulations. Companies should prioritize responsible AI development by collaborating with ethicists, following guidelines, making their AI models auditable, and making ethical considerations a core part of their decision-making processes. A culture of responsibility, not just technology, is key.
Could you introduce Swiss AI and explain the vision behind it? What led to its creation, and what role will it play in the evolving AI landscape?
Funded by EPFL and ETH Zurich, Swiss AI is a national initiative designed to advance the development of large-scale, open, and trustworthy foundation models for the benefit of society. It was created in response to the growing importance of AI technologies and the need for Europe—and Switzerland in particular—to play an active, independent role in shaping them. The initiative brings together researchers across institutions, providing shared access to large-scale computing infrastructure and fostering collaboration on ambitious, high-impact projects in areas such as healthcare, education, climate, and scientific discovery. By prioritizing openness, transparency, and societal alignment, Swiss AI aims to drive innovation while ensuring that AI systems are developed responsibly and remain accessible to academia, industry, and the broader public.
September 2025 marked the launch of Apertus. Could you tell us about Apertus, its capabilities, and what makes it different from its competitors?
Apertus is a large language model developed by the Swiss AI Initiative. With 70 billion parameters, it is the largest fully open and transparent model available. Designed for the public good, with regulatory compliance in mind, it is one of the first models built to fully comply with the EU AI Act. It has been trained with data in over 1000 languages. Apertus allows researchers, professionals, and enthusiasts to build upon the model and adapt it to their specific needs, as well as to inspect any part of the training process. This distinguishes it from models that make only selected components accessible.
There is growing discussion around technological sovereignty, especially when it comes to AI solutions developed in the United States or China. In your view, how important is digital sovereignty in AI? What needs to happen to make it a reality?
Digital sovereignty in AI is crucial for maintaining control over technology that could potentially influence society, national policies, economic interests, and individual freedoms. To achieve it, we need to invest in local research and development, establish regulations and guidelines, and promote cross-border collaboration. Switzerland, for instance, through projects like the Swiss AI Initiative or Apertus, aims to lead by example in ethical AI development.
AI is transforming the entire value chain, with companies investing heavily in areas such as operations, risk management, and customer service. How can organizations strike the right balance between pursuing productivity gains and maintaining proper control over risk?
Organizations should strike a balance by investing in AI for productivity while also implementing risk management frameworks, transparency mechanisms, and ethical guidelines. This involves setting clear goals that include both innovation and responsibility, monitoring AI outcomes, and ensuring human oversight where needed. The goal should be to augment workforce, not replace it.
What's coming next for Apertus?
Apertus will continue to evolve through updates, such as the 1.5 release this summer which will add multimodal inputs and better integration with tools. Our next developments are likely to involve a new generation of larger models, expanding the multilingual capabilities, responsibly integrating more data sources, and exploring new areas at the research frontier, together with other projects of the Swiss AI Initiative.
AI Community community
With Qorus memberships, you gain access to exclusive innovation best practices and tailored matchmaking opportunities with executives who share your challenges.