X Releases Open-Source Algorithm Amid Regulatory Scrutiny
Social media platform X has made its recommendation algorithm publicly available on GitHub, marking a renewed effort to address calls for greater transparency in content curation. The release, dated January 19, 2026, includes code for determining organic and advertising post recommendations in users’ “For You” feeds, fulfilling a commitment from owner Elon Musk to update the repository every four weeks with explanatory notes on modifications.
The system relies on a transformer model derived from xAI’s Grok architecture, which analyzes user engagement patterns to predict interactions such as likes, replies, or reposts. It processes in-network content from followed accounts alongside machine-learning-selected out-of-network posts, applying filters to exclude blocked or muted material, spam, or violent items. Scoring incorporates weighted probabilities for user actions, adjustments for content diversity, and author variety to avoid repetitive feeds.
This follows a partial algorithm disclosure in 2023, which faced criticism for lacking depth and completeness. Musk had then described the move as potentially “incredibly embarrassing at first” but essential for improving recommendations and building trust. The latest version represents a shift to a unified AI-driven approach, contrasting with the earlier rule-based system.
The timing coincides with heightened oversight. In December 2025, the European Union imposed a $140 million fine on X for breaching Digital Services Act transparency rules, particularly regarding its verification system. Additionally, X’s Grok chatbot has drawn intense criticism in early 2026 for generating nonconsensual sexualized images, including of women and minors, prompting investigations by UK regulator Ofcom, French authorities, and California’s Attorney General. Bans in Indonesia and Malaysia, along with restrictions limiting image generation to paid users, reflect global concerns over AI misuse.
