Apertus: The Fully Open Foundation Model for Sovereign AI

The Problem with "Open" AI

In 2026, the term "open source AI" has become increasingly misleading. Many models labeled as "open" release only their weights—the trained parameters—while keeping training data, code, and methodology hidden behind corporate veils. This creates significant problems for organizations that need true transparency: regulators who must verify compliance, researchers who need reproducibility, and enterprises that require auditable AI systems.

The EU AI Act demands documentation that most "open" models simply cannot provide. When you don't know what data trained your model, how can you prove it respects copyright opt-outs? When training code is proprietary, how do you verify safety properties? When alignment methods are secret, how do you audit for bias?

Enter Apertus: Fully Open by Design

The Swiss AI Initiative—a collaboration between EPFL, ETH Zurich, and CSCS—has released Apertus, a foundation model that takes "open" seriously. Everything is transparent: training data, code, weights, methods, and alignment principles. No black boxes, no hidden datasets, no proprietary secrets.

The name itself signals the philosophy: Apertus comes from Latin meaning "open" or "accessible." This isn't marketing language—it's a commitment to AI as a public good rather than a proprietary asset.

Why This Matters for Sovereign AI

Sovereign AI—the ability for nations and organizations to control their own AI infrastructure—has become a pressing concern. When your AI runs on models you can't inspect, trained on data you can't verify, running on infrastructure you don't control, you've surrendered significant sovereignty.

Apertus addresses this directly. Organizations can deploy it on their own infrastructure, audit every component, and modify it to meet specific requirements. For governments subject to the EU AI Act, for healthcare systems handling sensitive data, for financial institutions requiring regulatory compliance—Apertus offers something no proprietary model can: verifiable trust.

Technical Details

Apertus comes in two primary sizes: 8B and 70B parameters. Both are competitive with other open models at equivalent scales, but with a critical difference—you can verify every claim about performance, training, and safety.

The model is multilingual from day one, trained on over 1,000 languages. This isn't an afterthought or a fine-tune; multilingual capability was built into the foundation. For organizations operating globally, this represents a significant advantage over English-centric models retrofitted for other languages.

Compliance Built In

The EU AI Act isn't just regulatory burden—it's a framework for trustworthy AI. Apertus was designed with these requirements in mind: copyright opt-outs are respected, personally identifiable information (PII) is removed from training data, and the model includes safeguards against memorization of training content.

For organizations that have struggled to prove compliance with proprietary models, Apertus offers something unprecedented: documentation that actually satisfies regulatory requirements because the entire pipeline is transparent.

The Science of Open

Beyond compliance, open methodology enables scientific progress. When researchers can examine exactly how a model was trained, what data was used, and what alignment techniques were applied, they can build on that work meaningfully. The reproducibility crisis in AI research stems partly from proprietary training pipelines—Apertus offers a path forward.

The Swiss AI Initiative has documented alignment principles, making the values embedded in the model explicit rather than implicit. This transparency allows for informed debate about AI ethics rather than treating alignment as a black box.

Strategic Partnership and Availability

Swisscom serves as a strategic partner for the initiative, providing enterprise-grade infrastructure and deployment pathways. This combination of academic rigor and commercial viability positions Apertus for real-world adoption, not just research curiosity.

The model weights, training data, and code are available through standard channels. Organizations can sign up for the newsletter to stay informed about releases and updates.

What This Means for the Industry

Apertus represents a fork in the road for foundation models. One path continues with partially open models—weights available but methodology hidden—creating an industry dependent on vendor trust. The other path, exemplified by Apertus, treats AI as infrastructure that should be as transparent and auditable as any other critical system.

For enterprises, this choice will become increasingly important. As AI regulation intensifies and stakeholder demands for transparency grow, the hidden costs of opaque models—compliance overhead, audit limitations, vendor lock-in—will become more apparent. Apertus offers an alternative: a foundation model you can actually trust because you can actually verify.

The question isn't whether open AI will succeed—it's whether "open" will mean what it says. Apertus ensures it does.