Epsilon is a social network launched on September 28, 2026, presented as a European alternative to X (formerly Twitter). The platform is based on the open protocol ActivityPub and technically relies on Mastodon, which connects it to the Fediverse, the network of decentralized servers that can communicate with each other.
Epsilon and the ActivityPub Protocol: A Social Network Built on Mastodon
Contrary to what its communication suggests, Epsilon did not develop a technical infrastructure from scratch. The platform operates as a Mastodon instance, meaning a server that uses existing free software and a standardized communication protocol, ActivityPub.
This protocol allows an account created on Epsilon to interact with accounts hosted on other Mastodon instances or compatible platforms. In practice, an Epsilon user can follow and be followed by people registered on different servers, without those users needing to create an Epsilon account.
This technical choice has a direct consequence: the data published on Epsilon is not locked within a proprietary ecosystem. Unlike networks like Instagram or Threads, where content remains captive to the platform, the Fediverse relies on interoperability. To keep up with social media news and the alternatives emerging in France, googleplus.fr regularly offers analyses on these topics.
AI Moderation: Pre-Publication Filtering on Epsilon
Most social networks moderate content after it goes live, in response to reports. Epsilon takes a different approach: pre-filtering by artificial intelligence analyzes posts before they become visible.
The filtering tool is based on technology from Mistral, the French company specializing in language models. Human intervention is planned in case of doubt about automatic classification.
This system raises a technical question rarely addressed in public presentations: the false positive rate. A filter that is too strict may block legitimate content even before it is published. On a large scale, this type of preventive moderation can discourage some users, particularly those who post satirical or political content.
- The AI filter intervenes before publication, not after reporting
- The technology from Mistral powers the automatic classification system
- A human moderator takes over when the AI does not make a clear decision
Absence of Recommendation Algorithm: What It Means Practically
Epsilon displays posts in chronological order, without a personalized recommendation algorithm. On X, Instagram, or Threads, an algorithm selects and orders content based on each user’s past behavior: time spent on a post, interactions, inferred interests.
The absence of an addictive algorithm means there is no infinite feed optimized to increase time spent on the platform. Epsilon also does not offer autoplay for videos. The user chooses what to view, in the order in which the content was published.
This model has a cost in terms of user experience. Without an algorithm, the relevance of the feed entirely depends on the quality of the accounts followed. A user who follows few accounts will see a sparse feed, where an algorithm would have filled the gaps with suggestions. This operation favors users already familiar with decentralized networks, used to manually building their subscription feed.
European Hosting and GDPR Compliance: Epsilon’s Commitments
Epsilon positions itself as a social network fully hosted in Europe. User data remains stored on European servers, and the platform commits to strictly adhering to GDPR. No data is sold, and no advertising trackers are used.
This commitment distinguishes Epsilon from American platforms subject to the Cloud Act, a federal law that allows U.S. authorities to request access to data stored by U.S. companies, even when the servers are physically located in Europe.
The question of funding directly arises from this positioning. Without targeted advertising or data resale, Epsilon’s business model currently relies on the founders’ own funds and private investors. Paid offers for businesses and public figures are planned by the end of 2026, but no specific timeline has been communicated.
Adoption of Epsilon in France: Initial Figures to Put into Perspective
The media coverage surrounding Epsilon has been significant since its launch. However, the reported adoption figures remain modest and vary according to sources. Some mention around 1,000 active accounts at launch, while others report 5,000 registrations within the first four days.
This disparity illustrates a classic problem of new social networks: the difference between created accounts, registrations, and users who are actually active. A created account does not equate to a user who regularly posts or interacts.
For a social network to reach a critical mass, it must cross a threshold where the perceived value for each user justifies returning daily. Epsilon is not there yet. Compatibility with the Fediverse could theoretically accelerate this dynamic, as a new registrant immediately accesses a broader ecosystem of content than just the Epsilon user base.
Epsilon’s positioning remains a bet on the demand from a European audience weary of the practices of major platforms. Its medium-term viability will depend on the ability to transform media curiosity into regular usage and stabilize a business model without advertising.
