European Union (institutional) — European Union AI Act — first comprehensive AI regulation

2024

Artists: European Union (institutional) · Year: 2024 · Medium: Regulatory framework · Substrate: Regulation (EU) 2024/1689 — adopted 13 March 2024, entered into force 1 August 2024 · Place: Brussels, Belgium · Institution: European Parliament and Council The European Union AI Act was formally adopted by the European Parliament on 13 March 2024 and entered into force on 1 August 2024, establishing the world's first comprehensive regulatory framework for artificial intelligence systems. The Act's risk-based approach categorizes AI applications into four tiers: unacceptable risk (banned outright), high risk (subject to substantial compliance requirements), limited risk (transparency obligations), and minimal risk (largely unregulated). The Act's provisions apply to any AI system marketed or used within the EU, regardless of where the system is developed. For generative-AI image and video systems specifically, the Act introduces transparency requirements: AI-generated content must be marked or otherwise identifiable as AI-generated, with specific provisions for deepfakes that must clearly indicate their synthetic nature. Models trained on substantial volumes of copyrighted material are subject to disclosure requirements about training-data sources. The Act's general-purpose AI provisions (specifically targeted at large language models and large foundation models) are substantially binding on the major model developers (OpenAI, Anthropic, Google DeepMind, Stability, Meta) regardless of where they are headquartered. The Act's implementation has been phased through 2025 and 2026, with the most-substantive obligations on general-purpose AI taking effect August 2025 and the high-risk-system obligations taking effect August 2026. Compliance has been substantially expensive for major model developers, with several companies opening dedicated EU offices and substantially restructuring their operations to handle the regulatory burden. For artists working with generative AI, the Act's transparency provisions create new disclosure obligations for content distributed in the EU but largely do not constrain the kinds of work that can be made. Several artists in this archive have been involved in the AI Act's policy development, including Hito Steyerl (whose substantial public criticism of unrestrained AI development has been influential in EU policy circles). The Act's broader significance is regulatory: it establishes the EU as the first major jurisdiction with comprehensive AI regulation, and the resulting "Brussels effect" has substantially shaped subsequent AI regulation in other jurisdictions. The UK, US (state-level), Canada, and several Asian jurisdictions have been working on their own AI-regulatory frameworks substantially in dialogue with the EU model. Lore The EU AI Act had been in development since 2018 (the European Commission had announced AI policy intentions in 2018; the formal proposal came in April 2021). The Act's substantive negotiations during 2022-23 were substantially shaped by the post-ChatGPT generative-AI wave, which arrived during the legislative process and substantially expanded the Act's scope. Several major modifications to the Act during the trilogue negotiations of late 2023 specifically addressed the foundation-model provisions. The Act's final text was substantially more burdensome on US-based model developers than on European companies (which are generally smaller), prompting considerable corporate pushback during the implementation phase. Compliance investments by major developers have been substantial; OpenAI, Anthropic, and Google DeepMind have all opened or expanded EU offices since the Act's entry into force. Sources • European Commission — Regulation (EU) 2024/1689 (AI Act). • European Parliament — AI Act press archive.