Giotto admits that some of Alto and Solo’s model weights came from Kimi, raising questions about Europe’s sovereign AI claims

AT A GLANCE

Giotto accused of having its Alto model identify itself as Kimi，CEO admits that some weights came from Kimi, but the specific version and scope were not disclosed.

Article

Swiss Giotto claims to build Europe's strongest sovereign AI, admits after being debunked that “some model weights came from Kimi”

Beating AI News, Swiss AI laboratory Giotto released two models, Alto and Solo, claiming to have built “Europe’s smartest sovereign AI.” The official statement emphasized that the models are hosted on European cloud infrastructure, that data remains in Europe, and that they do not depend on US cloud service providers.

Problems were discovered by users immediately after the release. When testing Alto, one user found that the model actually identified itself as Kimi from China’s Moonshot AI. In response to questions from the community, Giotto co-founder and CEO Aldo Podestà admitted that the model had indeed inherited some Kimi weights, but stressed that the team had retrained it and added its own inference-time compute (TTC) technology. He did not say which version was used or exactly which weights had been inherited.

Even more ironically, Podestà published “Open Source Is a Trojan Horse” on October 19 and criticized Europe for packaging open-source models as “sovereign AI”, arguing that this often merely means depending on others in a different way. He did not expect to do the same himself just 8 days later—truly the pot calling the kettle black.

This is far from the first incident of its kind.September 2 Spanish AI company Multiverse Computing promoted Quasar 438B as “Europe’s leading AI model” without stating the model base in its release announcement. It was not until the next day that a technical article explicitly disclosed that it had actually been compressed and optimized from Zhipu’s GLM-5.2.

Original link https://m.theblockbeats.info/flash/374090

Key points

Swiss AI laboratory Giotto released two models, Alto and Solo, claiming to have built “Europe’s smartest sovereign AI”。
Giotto stated that the models are hosted on European cloud infrastructure, that data remains in Europe, and that they do not depend on US cloud service providers.
The article says that, while testing Alto, someone found that the model identified itself as Moonshot AI’s Kimi。
The article says that Giotto co-founder and CEO Aldo Podestà admitted that the model had inherited some Kimi weights and said that the team had retrained it and added its own inference-time compute (TTC) technology.
The article says that Giotto did not specify which Kimi version was used or which specific weights had been inherited.
The article says that Aldo Podestà published “Open Source Is a Trojan Horse” on October 19 and criticized Europe for packaging open-source models as “sovereign AI”。
The article says that Multiverse Computing September 2 promoted Quasar 438B as “Europe’s leading AI model”, and disclosed the next day in a technical article that it was based on Zhipu’s GLM-5.2 and had been compressed and optimized.

AI-assisted interpretation

The following is analysis, separate from reported facts. Verify important claims independently.

“sovereign AI” in this report emphasizes model hosting, data storage, and dependence on external service providers. The point of contention is that Giotto emphasized European hosting and independence from US cloud service providers, but its CEO was also reported to have admitted that the model partially inherited Kimi weights. European deployment and fully autonomous model weights are not the same thing.

Why it matters to readers

The incident has subjected Giotto’s “sovereign AI” claims to scrutiny over transparency, and also shows that when assessing a model’s autonomy, it is not enough to look only at where the servers and data are located; the source of the model weights and whether the training process has been adequately disclosed must also be considered.

For newcomers, it is useful to distinguish three things: where the model runs, where the data is stored, and where the model weights come from. The article gives Giotto’s official position only on the first two points; the specific version and scope of the weight source remain unclear.

Risks and unknowns

The evidence is a media brief paraphrase, without an attached direct statement from Giotto or Aldo Podestà acknowledging use of Kimi the weights.
It is unclear whether both Alto and Solo inherited the Kimi weights; the original wording is insufficiently clear.
The specific version used and the scope of the inherited weights have not been disclosed.Kimi

Giotto admits that some of Alto and Solo’s model weights came from Kimi, raising questions about Europe’s sovereign AI claims

Article

AI-assisted interpretation

Why it matters to readers

Risks and unknowns

Related Developments

Related concepts

Keep following

Security Center