EU AI rules trigger industry debate

Within the breakneck global AI race, Switzerland has announced the launch of Apertus, its first fully open-source large language model (LLM), developed in partnership between ETH Zurich, EPFL, and the Swiss National Supercomputing Centre. The model has been designed to be freely accessible to the public, with its code, training data, and technical details openly available under an Apache 2.0 license. This model is presented as a trustworthy and transparent alternative at a time when many major AI systems face increasing scrutiny over bias, data usage, and legal compliance.

One of the most distinctive features of Apertus is its commitment to multilingual inclusivity. Unlike many existing large language models, which are dominated by English training data, Apertus was trained on material covering more than 1,000 languages. This includes not only major global languages but also underrepresented ones like Swiss German and Romansh. According to the claim, around 40% of the training dataset comes from non-English sources, making the model far more diverse in its linguistic coverage than many of its commercial counterparts.

The project was made possible through the Alps supercomputer at the Swiss National Supercomputing Centre in Lugano. This system provided the computational power needed to train the model at scale while also ensuring that the process was powered by carbon-neutral energy sources. The result is a model released in two forms – one version with eight billion parameters designed for lighter use cases, and a more powerful seventy-billion parameter version suited for research and enterprise applications.

The creators of Apertus also made ethics and compliance central to the project. Data collection respected website rules like robots.txt, avoiding unauthorized scraping of restricted content. The model was built to align with Swiss and European regulatory standards, including the EU AI Act.

Apertus is already being made available through multiple channels. Developers and institutions can access it via the Swisscom Swiss AI Platform or download it directly from Hugging Face (the online hub for open-source AI models). Even to encourage experimentation and community input, the model will also play a central role in Swiss {ai} Weeks, a series of hackathons and workshops where developers, businesses, and students will be invited to test its capabilities and contribute to its evolution.

Importantly, the research teams plan to update Apertus regularly and develop customized versions for fields like healthcare, education, law, and climate science. While it may not be as large as proprietary models like OpenAI’s GPT-5 or Anthropic’s Claude, its creators argue that the importance of a model goes beyond scale. According to them, Apertus stands out for its full transparency, legal compliance, and adaptability (qualities many commercial systems lack). They also argue that Apertus should not be viewed as a commercial product but as a form of public infrastructure, comparable to roads, electricity, or telecommunications networks.

The Tech Portal is published by Blue Box Media Private Limited. Our investors have no influence over our reporting. Read our full Ownership and Funding Disclosure →