Switzerland's Giotto claimed to build Europe's strongest sovereign AI, but after being exposed, admitted that 'some model weights come from Kimi'.

Beating AI News Flash: Swiss AI lab Giotto has released two models, Alto and Solo, claiming to have built "Europe's smartest sovereign AI." The company emphasized that the models are hosted on European cloud infrastructure, with data remaining in Europe and no reliance on US cloud providers.

But problems were spotted by netizens right after the launch. When someone tested Alto, the model claimed to be Kimi from China's Moonshot AI. Faced with community scrutiny, Giotto co-founder and CEO Aldo Podestà admitted that the model did inherit some Kimi weights, but stressed that the team retrained it and added its own test-time compute (TTC) technology. As for which version was used and exactly which weights were inherited, he said not a word.

More ironically, Podestà published "Open Source Is a Trojan Horse" on October 19, criticizing Europe for packaging open-source models as "sovereign AI" when it is often just another way of depending on others. Just 8 days later, he did the same thing himself—truly the pot calling the kettle black.

This kind of thing has happened more than once. On September 2, Spanish AI company Multiverse Computing marketed Quasar 438B as "Europe's leading AI model" without disclosing the model's base in the launch announcement. Only the next day did it clarify in a technical article that it was actually built by compressing and optimizing Zhipu's GLM-5.2.