EU regulators wanted to see inside X. Musk now wants users to see how governments intervene in what appears on the platform.
Elon Musk with former EU Internal Market Commissioner Thierry Breton, who oversaw the introduction of the Digital Services Act during his time in Brussels. Photo: Twitter/Thierry Breton
|
The standoff between Elon Musk and Brussels over the regulation of X is entering a new phase. The European Commission has made algorithmic transparency one of the pillars of the Digital Services Act (DSA) in order to understand how major platforms select, rank and recommend content. Musk has chosen to take Brussels at its word: rather than simply complying with EU requirements, he intends to lift the lid on X and, in the process, reveal who is trying to influence what appears on the platform.
This move is not entirely new. In March 2023, after Musk’s acquisition, Twitter published part of the code used to power the recommendations in its feed. But that initial initiative remained limited in scope. The real change came in January 2026, when X released the code for its For You feed recommendation system. The code pertains to the core mechanism that determines which posts are suggested to users.
From Compliance to Exposure
What X is making public is therefore not its entire algorithm, nor does it make the platform’s whole operation transparent. More specifically, X has published the mechanism that selects and ranks posts likely to interest each user. The repository is publicly accessible on GitHub and published under the Apache 2.0 license.
The system retrieves posts from two sources: accounts the user follows and posts from accounts the user does not follow, selected by a machine learning system. This content is then ranked by Phoenix, a model developed using technology from Grok, the artificial intelligence developed by Musk's xAI. The model does not simply assess whether a post is likely to appeal to a user; it estimates the probability that the user will like, share, reply to, quote or take other actions on it. The final ranking then depends on these predictions.
In other words, X is now revealing a substantial part of the mechanism that determines what its users see – and, above all, what is likely to be amplified or demoted in their feed. The system is even designed to learn from each user’s engagement history, rather than relying solely on a set of manually written rules.
On 15 May 2026, X went further. The new version released on GitHub includes an end-to-end inference pipeline that reproduces the entire process from content retrieval to ranking, as well as new components dedicated to content understanding, advertising and candidate selection.
Musk’s answer to the European regulatory approach is technical transparency. Making the algorithm’s inner workings public also raises another question: what happens when content that should appear in the feed is subject to external intervention? His strategy is to turn the question around: if regulators demand to know how the algorithm works, why not also make public the requests sent to the platform by government authorities?
Transparency Cuts Both Ways
In its latest transparency report, X reported receiving 3,831 content-removal requests from authorities in EU member states, acting on 3,464 of them (90.42%). It also received 8,730 information requests about accounts from the same sources and responded to 47.03% of them.
These requests come almost entirely from national courts, police, prosecutors and other designated bodies in individual EU countries – not from the European Commission. While the Commission has broad powers to investigate platforms, demand information, find non-compliance and impose large fines, it does not issue the individual content-removal orders recorded under the EU heading in X’s report.
The figures are striking, and that is precisely the intended effect. But those numbers need some context. The exact content of the requests is not yet known. A takedown request is not necessarily a request for political censorship but may involve the enforcement of a court order or content that is actually illegal under the law of a member state. X also notes that it reviews the legal validity of requests and may challenge those it deems excessively broad or legally flawed. Some requests may simply seek to restrict content geographically.
But true to form, Musk intends to turn that visibility to his advantage.
While member state authorities issue the specific removal requests, the Commission sets the rules, polices compliance at the platform-wide level, and can punish companies for what it judges to be inadequate overall moderation. Since the start of his conflict with the EU, Musk has repeatedly criticized these measures as excessive influence from Brussels over content moderation.
In December 2025, the European Commission fined X €120m ($140m) under the DSA, citing the blue-checkmark system, the lack of transparency in the advertising registry and insufficient access for researchers to public data. X’s response went well beyond addressing those concerns.
By publishing its code, X does not claim to have made its entire platform perfectly transparent, but it allows researchers, developers and the public to examine part of the system that governs content visibility. And by simultaneously revealing the requests made by public authorities, Musk has exposed another layer of the system: external interventions capable of altering what users can see. In doing so, he transforms a regulatory obligation into a political tool. Transparency thus ceases to be a one-way mechanism imposed by the regulator and becomes a double-edged sword.
It is a brilliant move that EU institutions as well as member state authorities may find difficult to counter. The stakes are high, as Germany and France are approaching high-profile elections and have already signaled their determination to combat attempts to interfere with the vote. Voters will now have a clearer view of how national authorities and EU regulators seek to shape the online information environment in which elections are fought.
