Independent EU AI Act information resourceLegal text verified against EUR-Lex
Menu
Check readiness

MEPs extend ban on social scoring

What was reported

The European Parliament has voted to extend the ban on social scoring systems to private companies and reduce the role of the European AI Office. This includes changes such as removing references to the principles of data minimisation and data protection by default, excluding open-source AI models from the scope of the regulation and introducing regulatory sandboxes to help developers of high-risk AI systems.

What has happened since this was written

This piece was published on 28 February 2023, while the Regulation was still moving through the EU legislative process, so it describes a draft. Everything in the table below has happened since. Any article number, penalty figure or date in the text above should be read against the adopted and amended Regulation rather than relied on.

The Regulation’s progress since 28 February 2023
DateWhat happenedStatus
1 Aug 2024The Regulation enters into forceIn force
2 Feb 2025Prohibited practices and AI literacy applyIn force
2 Aug 2025General-purpose AI model obligations applyIn force
2 Aug 2026Transparency obligations and the general application dateIn force
2 Dec 2026Two new prohibitions and the legacy marking deadlineUpcoming
2 Dec 2027Annex III high-risk obligations applyUpcoming
2 Aug 2028Annex I product-safety high-risk obligations applyUpcoming
2 Aug 2030Legacy large-scale IT systems must complyUpcoming

The text was also amended by Regulation (EU) 2026/1744, the Digital Omnibus on AI, which deferred the high-risk deadlines, added two prohibited practices, softened the Article 4 AI literacy duty and gave the AI Office direct enforcement powers.

For the position under the current text, start with the implementation timeline, the compliance checklist or our analysis. The full legal text is at the Act explorer.