EU AI Act brings new transparency duties for deepfakes

Since 2 August 2026, certain AI systems in the EU must be designed so people can understand when they are interacting with AI or viewing AI-generated content. Article 50 creates transparency requirements covering AI interactions, synthetic content identification and some deepfake disclosures, including obligations around machine-readable markings and clear notices when deepfakes are involved.

Businesses using general-purpose AI to generate synthetic audio, image, video or text content, or to assess emotions or biometric data, may need to review their responsibilities. Providers are expected to build watermarking or metadata into generation pipelines, while deployers may need visible labels or disclosures when publishing realistic deepfakes or text on matters of public interest.

The rules are intended to reduce deception, manipulation, fraud, impersonation and consumer confusion, although they do not prohibit fraud directly. UK businesses may still face practical pressure to align with EU standards, particularly when trading internationally, and failures to meet transparency and watermarking mandates can lead to fines of up to €15 million or 3% of a company’s total annual global turnover.