We’re excited to announce EuroGPT, an open source and privacy conscious alternative to ChatGPT for businesses.

Today, businesses in Europe face two challenges when leveraging AI applications. First, they need to use proprietary AI chat applications that offer limited transparency or control. Second, their data gets transferred to the US, making them lose jurisdiction over this data.

EuroGPT is designed to overcome these challenges and more. It’s built completely as an open source alternative, runs open weight LLMs in Germany, and enables enterprises to keep their knowledge bases local to their environments.

Businesses can start evaluating EuroGPT at https://eurogpt.ubicloud.com.

Garry Tan, the President and CEO of Y Combinator, shared his enthusiasm by saying, “Innovation thrives on transparency and collaboration. What EuroGPT is doing -combining open-source infrastructure with the flexibility to control your own data- is exactly what we need in today’s AI landscape.”

How EuroGPT works

EuroGPT Enterprise combines open source and open weight solutions. These include Ubicloud’s open source and portable cloud, Meta’s Llama 3.1 405B open model, Llama Guard for input and output moderation, and DuckDuckGo for privacy protected web searches. The following architecture diagram offers a simplified view.

The solution also offers robust data privacy and security controls. It complies with GDPR and ensures that all data remains within the EU. Further, EuroGPT doesn’t keep any customer data or metadata for training purposes; and offers custom data retention windows.

EuroGPT encrypts data in transit and at rest. The encryption keys are regularly rotated; and we also apply envelope encryption for data at rest. This ensures that compromising either the storage host or the key repository alone isn’t sufficient to decrypt the data. Ubicloud is also happy to provide SOC 2 audit or recent penetration test results to interested parties. Finally, customers can validate our privacy and security controls thanks to using a fully open source solution.

Improve your AI results with RAG. Keep your knowledge base private

Retrieval-augmented generation (RAG) enables LLMs to access and use external information sources dynamically when generating responses. This way, organizations can integrate their documents or knowledge bases with AI. They can then get highly specialized, accurate, and context-aware responses.

One way to use RAG with AI chat applications is to send your knowledge base over the public infrastructure to an Embeddings API. Once you generate embeddings, you’d then put them into a managed vector database. For enterprises with sensitive data however, offloading your knowledge base to public infrastructure has security and compliance implications.

EuroGPT allows enterprises to keep their knowledge bases local to their environments. Again, this is thanks to the flexibility that comes with open source. You can find a simplified view of this feature below. Please note that the feature is only available through contract with a minimum commitment.

Finally, we believe that these privacy and security features should be available to all businesses in Europe. EuroGPT is priced at €19 per user per month, at 3x lower cost than comparable solutions in the US. We also provide contracts tailored towards organizational needs.

In summary, EuroGPT Enterprise is the first fully open source solution that gives you four distinct benefits. These are:

• transparency and control,
• over data that stays in Europe,
• with the option to keep your knowledge base private,
  
• and at 1/3rd price of comparable solutions.
  

For more information, please contact us at [email protected].