What is Eustella? The vision of newsrooms.ai
Eustella aims to position itself as a European alternative to ChatGPT, Claude and Gemini. The company behind the service is newsrooms.ai, or AI Newsrooms Technology GmbH, based in Vienna, Austria.
The company's core objective is to create an AI infrastructure that adheres to European values of privacy, data sovereignty and control. Currently, the service is in open beta, but the LLM (Large Life Management) platform is already accessible via a web browser (link) and via the app (available on the App Store and Play Store) for iOS and Android.
Data protection as a foundation: Hosting on German IONOS servers
Eustella is aimed at anyone who wants to use chatbots, agents and similar services just like with the major providers, but without having to share their personal data. With Eustella, users don't simply feed their data to a European provider; rather, the company claims to collect virtually no data and does not engage in profiling. The AI is hosted on servers belonging to German company IONOS, which has no American parent company. IONOS cannot view the encrypted chat content.
The technology behind AI: Open-source models instead of proprietary training
Since the website or app hosts one or more open-source (OS) models, newsrooms.ai doesn't rely on its own AI training but instead uses the most modern and promising OS LLMs. This means the provider doesn't need to use any personal data for AI training. The only data collected is for product improvement (e.g., loading times, error rates, latency, clicks).
There is, in fact, one tool from a US provider in the data processing chain: the PostHog software from the American provider of the same name is used for product analysis. However, it is configured to run only on servers in Frankfurt and only analyzes technical metadata (clicks, loading times), but not chat content.
Pricing, whether there will be a free tier as usual, and how limited such access will be are currently unknown, as newsrooms.ai intends to evaluate the experiences from the beta phase first.
Outlook: How does open-source AI perform in practical testing?
The sole use of open-source models could, of course, lead to a loss of quality in the answers compared to those from major providers. I tested whether this is the case in my next article: I had a travel agent generate tips for a two-day trip to Paris, planned my DSA (The Dark Eye) evening, and coded a small tool using Vibe.












