By Stuart Kerr, Technology Correspondent
Published: 14/09/2025 | Updated: 14/09/2025
Contact: [email protected] | @LiveAIWire
A Sovereign Shift in AI
The global AI race has been dominated by American and Chinese companies, with Europe often cast as a regulator rather than an innovator. That narrative shifted this month when Switzerland announced Apertus, a large language model released under a fully open licence. As The Verge reported, Apertus is not only free to use but comes with openly published weights, architecture, and training data — a level of transparency unprecedented in the field.
Apertus: Built for Openness
The launch was backed by EPFL, ETH Zurich, and the Swiss National Supercomputing Centre. According to the EPFL/ETH press release, Apertus comes in 8-billion and 70-billion parameter versions, supports more than 1,800 languages, and is hosted on Hugging Face and Swisscom infrastructure. Researchers worldwide can audit the training data, replicate benchmarks, and adapt the model without hidden barriers.
Technology analysts are calling it a sovereign European alternative to US-developed models. InfoWorld framed the release as an “ethical counterweight” to closed models like GPT-4 or Gemini, aligning with Europe’s priorities of privacy, accountability, and inclusivity.
Building Trust
Trust has become central to the AI conversation. Proprietary systems face mounting criticism for opacity and bias, while transparency has emerged as a differentiator. Swissinfo.ch described Apertus as a “transparent ChatGPT alternative,” positioning it as digital public infrastructure rather than a corporate product.
This resonates with debates we covered in Beyond Algorithms — Hidden Carbon & Water, where hidden resource costs of AI became visible only after scrutiny. Just as sustainability demands full disclosure of carbon and water usage, accountability in AI depends on companies being open about what powers their models.
Implications for Generative AI
The openness of Apertus challenges assumptions that cutting-edge AI must remain closed. A review on OpenDataScience.com highlighted that Apertus is already being used by startups, universities, and even government agencies, proving that transparency can coexist with practical deployment.
It also raises questions about how value flows through digital ecosystems. In Can Publishers Survive Zero-Click Era?, we argued that AI-driven summaries threaten publishers’ revenue streams by cutting out the middleman. Apertus, by contrast, redistributes value in a different way — taking power away from AI gatekeepers and placing it in the hands of the broader research community.
Risks and Responsibilities
Of course, openness introduces its own risks. Security researchers caution that open weights can be exploited for disinformation or cyberattacks. Critics ask whether making powerful models available without restrictions could worsen harms. But proponents argue that open-source scrutiny actually improves safety. By enabling third parties to audit and stress-test the system, Apertus may avoid some of the blind spots plaguing proprietary models.
This tension echoes themes from AI and Emotional Manipulation, where we explored how technology’s social consequences are often tied less to the algorithms themselves and more to the governance structures around them. Apertus will test whether a community-driven governance model can succeed where corporate secrecy has failed.
Europe’s Defining AI Moment
The symbolic weight of Apertus is hard to overstate. By releasing a fully open, multilingual AI system, Switzerland has shown Europe is not just a referee in the AI game but a player with its own vision. The model could set a precedent for other nations to pursue openness as a form of sovereignty, ensuring AI reflects diverse cultural and social values rather than just those of Silicon Valley or Beijing.
Whether Apertus can compete in scale with proprietary giants remains uncertain. But as InfoWorld noted, the question may no longer be about scale alone but about legitimacy. And legitimacy, in a digital era rife with scepticism, may be the real differentiator.
About the Author
Stuart Kerr is the Technology Correspondent for LiveAIWire. He writes about artificial intelligence, ethics, and how technology is reshaping everyday life. Read more