EU rules for general-purpose AI: what applies from when, and who has signed the Code of Practice

What we observe

Under Article 113 of Regulation (EU) 2024/1689, Chapter V, which sets the obligations for providers of general-purpose AI models, applies from 2 August 2025, with the exception of Article 101 on fines [1]. The Regulation's general date of application is 2 August 2026, with staged exceptions [1, 4]. Providers of general-purpose AI models placed on the market before 2 August 2025 must take the necessary steps to comply by 2 August 2027 (Article 111(3)) [1].

A model is presumed to have high-impact capabilities, and so systemic risk, when the cumulative computation used for its training is greater than 10²⁵ floating point operations (Article 51(2)) [1]. Under Article 101, the Commission may fine providers of general-purpose AI models up to 3% of annual total worldwide turnover in the preceding financial year or EUR 15 000 000, whichever is higher [1]. The Commission states that its enforcement powers enter into application from 2 August 2026, when it will enforce the obligations, including with fines [3].

The Commission says the Code of Practice was published on 10 July 2025, and that the Commission and the AI Board have confirmed it is an adequate voluntary tool for providers to demonstrate compliance [2]. The Code has three chapters: Transparency, Copyright, and Safety and Security; the last is relevant only to providers of models with systemic risk under Article 55 [2].

On 26 Oct 2026 the Commission's page lists 22 signatories: AI Studio Delta; Aleph Alpha; Almawave; Amazon; Anthropic; Black Forest Labs; Bria AI; Cohere; Domyn; Dweve; Fastweb; Google; IBM; LINAGORA; Microsoft; Mistral AI; Open Hippo; OpenAI; Pleias; Reflection AI, Inc.; ServiceNow; and WRITER [2]. The page adds that xAI signed up to the Safety and Security Chapter only, and so will have to demonstrate compliance with the transparency and copyright obligations via alternative adequate means [2]. The Commission notes that some signatories may not appear immediately [2].

Why it matters

The timetable is staged. New models have faced the obligations since August 2025; fines become available a year later; models already on the market before August 2025 have a further year. Signing the Code is voluntary, and the Commission describes it as a way to show compliance with less administrative burden [2]. A partial signature, as with xAI, leaves the other obligations to be met by other means [2]. The signatory list changes over time; this count is as at 26 Oct 2026.

Method

Sources: the Regulation as published in the Official Journal on 12 Jul 2024, read on EUR-Lex [1], and two Commission pages [2, 3], all accessed on 26 Oct 2026. The text was checked against the amendments made by Regulation (EU) 2026/1744 (Digital Omnibus on AI), published in the Official Journal on 24 Jul 2026 and in force since 27 Jul 2026 [4]; it does not amend the provisions cited here. Signatories are counted as listed under "Signatories of the code of practice"; xAI is reported separately, as the page does. We do not name non-signers, since the page does not. Framing follows AIO Watch method note AIO-M01, v1.2 [5].

References

1. [1] European Parliament and Council. Regulation (EU) 2024/1689 (Artificial Intelligence Act). OJ L, 12 Jul 2024. https://eur-lex.europa.eu/eli/reg/2024/1689/oj. Accessed 26 Oct 2026.
2. [2] European Commission. "The General-Purpose AI Code of Practice". https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai. Accessed 26 Oct 2026.
3. [3] European Commission. "Guidelines for providers of general-purpose AI models". https://digital-strategy.ec.europa.eu/en/policies/guidelines-gpai-providers. Accessed 26 Oct 2026.
4. [4] European Parliament and Council. Regulation (EU) 2026/1744 (Digital Omnibus on AI), amending Regulation (EU) 2024/1689. OJ L, 24 Jul 2026. https://eur-lex.europa.eu/eli/reg/2026/1744/oj. Accessed 26 Oct 2026.
5. [5] AIO Research Team. AIO Watch method note AIO-M01, v1.2, 25 Oct 2026.