Responsible AI tools with a focus on European data storage, sustainability, and privacy

Which AI tool is right for me?

When choosing a generative AI tool, it is important to consider questions such as: does the tool align with my AI policy (opens in new tab)? What are my ambitions regarding sustainability and privacy? How important is it that my data is not used to train AI models? And do I want data to be stored on European servers?

But also: how high does the quality of the output of an AI tool need to be? Most mainstream AI tools (ChatGPT, Gemini, Copilot) often deliver higher quality than more responsible AI tools. So, an essential question that you, as a user or organization, must ask is: am I willing to trade off some quality for more responsible use?

Six more responsible generative AI tools

Mistral: European data and digital sovereignty

A key difference between mainstream AI tools and European alternatives lies in data storage. The French company Mistral (opens in new tab) develops open-source language models within European data and privacy regulations. Mistral’s Vibe (opens in new tab) (formerly Le Chat) works similarly to ChatGPT or Gemini. For cultural institutions, this means more control over where data is stored and how data is used.  

The paid version of Microsoft’s Copilot also offers data storage on European servers.  

GreenPT: Sustainability and privacy first

AI is energy-intensive. GreenPT (opens in new tab) is a Dutch initiative that makes generative AI available via servers running on sustainable energy. GreenPT uses open-source models from other providers and refines them for specific tasks, such as writing and reasoning. The approach complies with European privacy regulations, and data remains within Europe.

Lumo: Privacy as a core value

Those who know Proton (opens in new tab) from its encrypted email services and VPN services know that this company values privacy as a core principle. The recently launched AI assistant Lumo (opens in new tab) operates on the same principle. Lumo does not collect data, does not train models with user input, and uses open-source language models on European servers.

Public AI and Apertus: open Swiss AI

Apertus (opens in new tab) is a free, fully open AI model from Switzerland. Through Public AI (opens in new tab), you can use Apertus (opens in new tab) for free. The construction, training data, and operation are public. The model supports more than a thousand languages, including smaller languages often overlooked by other models. Another way to use Apertus is as a local AI model. Learn how to do that here. (opens in new tab)  

Thaura AI: an activist alternative

Thaura (opens in new tab) positions itself as an activist alternative to major American AI providers. The tool uses open-source models from other providers, does not train on user data, shares nothing with third parties, and uses less energy than well-known models. Thaura also aims to amplify non-dominant voices in societal issues. An AI with a distinct ethical and political profile.  

EU GPT: European AI for organizations

EU GPT (opens in new tab) is also a Dutch AI tool for organizations that want to handle data carefully. The tool offers many different open-source models running on European servers. Data is never used for training.  

Other responsible AI initiatives

In addition to these AI tools, there are other interesting developments contributing to responsible AI usage. We continue to monitor the progress of these initiatives so you don’t have to.

• Hugging Face (opens in new tab): a platform that provides open-source AI tools and large language models. This allows you to experiment freely while keeping your data safe. Learn how to work with local AI models in this article (opens in new tab).
• LibreChat (opens in new tab): on this open-source platform, you can work on all AI tools you’re subscribed to in one place. No more switching between browsers.
• GPT-NL (opens in new tab): a Dutch initiative, supported by the government and knowledge institutions SURF and TNO, working on its own language model. The goal: an open, accessible, and responsible infrastructure for Dutch language and context. Currently, pilots are being conducted to test the model. 
• EuroLLM (opens in new tab) and OpenEuroLLM (opens in new tab): a collaboration of European universities and companies to develop language models that can form the basis for new European AI tools. Based on European values: open, diverse, and compliant with European legislation.

The first more responsible generative AI tools are now available, and more are expected to follow. What choice will your organization make?