Partially EffectiveEffective: Aug 2 deployer transparency deadline holds; HRAIS obligations extend to Dec 2 2027 / Aug 2 2028 · Updated Jul 20, 2026
Regulatory content changes frequently. This page is informational only and should be verified with qualified counsel before reliance.
Applies to
Providers and deployers of AI systems serving EU audiences, including those generating or altering public-facing content (e.g. synthetic media, deepfakes).
Business impact
Agencies serving EU brands or audiences should not assume the obligations are purely the vendor's. Deployer-side transparency expectations apply now.
Disclosure relevance
If your agency produces public-facing AI-generated content for EU audiences, you need a repeatable process for labeling, review, and documentation.
Metadata relevance
Watermarking and content-provenance metadata is explicitly in scope for certain generated media.
Key obligations
▸Transparency obligations for AI systems and AI-generated or altered content under Article 50
▸Aug 2 deployer transparency deadline applies
▸The Dec 2 grandfather clause is vendor-only (watermarking obligations)
▸High-Risk AI System (HRAIS) obligations extended to Dec 2 2027 / Aug 2 2028 (per May 2026 Omnibus)
How Atlas² helps
We help build a labeling + documentation process for AI-generated media and map which obligations fall on you versus your tooling vendors.
We use cookies that are necessary to run the site. With your consent we also use analytics and advertising cookies. You can change this anytime under Your Privacy Choices. See our Privacy Policy.