EU publishes transparency guidelines for AI systems ahead of August deadline

The European Commission has published a set of recommendations to help AI providers comply with transparency obligations under the EU's AI Act, which take effect on August 2, 2026. The rules require that users be informed when interacting with an AI system, such as chatbots or voice assistants, and that AI-generated content (text, images, audio, video) be clearly marked with watermarks to indicate its artificial origin Source: infobae.

Scope and exceptions

The obligations apply to providers and deployers of AI systems used in public services, customer service, fraud reporting, and other contexts Source: diariolibree. However, the watermark requirement for most content has been delayed until December 2, 2026, except for "deepfakes" involving sensitive information like elections, public health, or political events, which must be labeled from August 2 Source: elpaises. Artistic, satirical, and fiction content are exempt. Additionally, emotion recognition and biometric categorization systems must inform users, unless authorized by law for crime prevention Source: infobae.

Challenges for businesses

While the recommendations are voluntary, companies that do not follow them must demonstrate compliance with the binding transparency rules. Justo Hidalgo, senior director at Adigital, notes that businesses face uncertainty due to a lack of harmonized standards and definitive technical guides, particularly affecting SMEs Source: abces. The European Commission has also introduced the 'Digital Omnibus' to delay some high-risk AI obligations until 2027-2028 Source: abces. Penalties for non-compliance can reach up to 3% of annual turnover or €15 million, with lower fines for EU institutions .