What Took Effect
<cite index="3-1,3-2">On 20 July 2026, the European Commission adopted guidelines on the transparency obligations under Article 50 of the EU Artificial Intelligence (AI) Act, which apply from 2 August 2026. These obligations require providers and deployers of AI systems to be transparent about the use of AI in four key areas: direct interaction with individuals; AI-generated content; emotion recognition and biometric categorisation; and deepfakes and AI-generated text on matters of public interest.</cite>
<cite index="16-5,16-6">Article 50 of Regulation (EU) 2024/1689 establishes transparency obligations for providers and deployers of specific categories of AI system, and is the cornerstone provision of Title IV, addressing the risk that individuals may be deceived about whether they are interacting with a human or an AI system, or whether content they consume has been artificially generated.</cite>
Key Obligations
<cite index="16-7">Under Article 50(1), providers of AI systems designed to interact directly with natural persons must ensure those systems are designed in a way that allows users to be informed, clearly and without ambiguity, that they are interacting with an AI system.</cite> <cite index="16-9">Providers of general-purpose AI (GPAI) models generating synthetic content must implement technical solutions — such as watermarking or metadata tagging — to ensure outputs are identifiable as AI-generated and machine-detectable by downstream systems.</cite> <cite index="12-1">Deployers of an AI system that generates or manipulates image, audio or video content constituting a deepfake are required to disclose that the content has been artificially generated or manipulated.</cite>
A limited transition carve-out applies in one area: <cite index="1-15">the AI Omnibus provisional agreement of May 2026 grants generative AI systems already on the market before August 2, 2026, until 2 December 2026 to meet the machine-readable marking requirement.</cite>
Enforcement Architecture
<cite index="7-1,7-2">Compliance will mainly be enforced by national competent market surveillance authorities. The AI Office has a more limited role, with jurisdiction only over AI systems built on GPAI models where the same entity provides both the system and the model, or where the AI system is integrated into a very large online platform designated under the Digital Services Act.</cite>
<cite index="21-4">The AI Office assembled a 38-person enforcement team in Brussels authorised to interview company staff, access model documentation, and revoke EU market access for non-compliant providers.</cite>
<cite index="3-3">Non-compliance can attract fines of up to €15 million or 3% of worldwide annual turnover.</cite>
Impact on Frontier Model Providers
<cite index="23-3,23-4">The enforcement powers, which took effect August 2, are administered by the Commission's AI Office and cover every company that makes a GPAI model available in the EU, no matter where that company is headquartered. That puts US companies including Anthropic, OpenAI (maker of ChatGPT), and Google among those subject to the new rules.</cite>
<cite index="19-4,19-5">The obligations fall on deployers as well as providers — and a foundation model vendor's EU compliance documentation covers the vendor's obligations, not those of downstream API builders who surface model outputs to end users.</cite>
<cite index="21-2,21-3">More than 180 organisations — including Anthropic, Google, Microsoft, OpenAI, Amazon, IBM, Mistral AI, and Cohere — have signed the Code of Practice on GPAI transparency, which confers regulatory deference in enforcement proceedings. Meta declined to sign, citing legal uncertainty, and must demonstrate compliance through alternative means.</cite>
<cite index="8-6,8-7,8-8">Signing on to the Code of Practice on Transparency of AI-generated Content is one recognised compliance path. Organisations that decline must demonstrate compliance through alternative means the Commission considers adequate, though what those alternatives look like in practice is not spelled out in detail, with that judgement falling to national market surveillance authorities.</cite>
<cite index="20-9,20-10">For model providers such as Anthropic and OpenAI, entering the EU market now requires increased compliance resources, including risk assessments, documentation, and incident notification mechanisms — costs that may divert some research and development funding and potentially delay advanced model launches in Europe in the near term.</cite>