EU AI Act transparency rules begin, high-risk obligations delayed to

On August 2, 2026, the next phase of the European Union’s Artificial Intelligence Act came into force, introducing new transparency requirements for AI systems. The rules require chatbots and other AI that interact directly with people to disclose their artificial nature, and mandate that synthetic content be machine-readable, while also obligating emotion recognition and biometric categorization systems to inform individuals of their use. Breaches can lead to fines of up to 15 million euros or 3 percent of global annual turnover, whichever is higher.

Companies face operational rather than transformational changes, as the provisions add a compliance layer without requiring prior approval or abandonment of AI systems. Organizations must now identify where AI is embedded in products, customer interactions, and internal processes—including third-party tools—and ensure those systems meet the new transparency standards.

The Act’s most far-reaching requirements for high-risk AI systems, originally scheduled for August 2, 2026, were postponed to December 2, 2027, as part of the Digital Omnibus package agreed in May. These obligations cover areas such as biometrics, employment, education, essential services, and migration, and include risk management, documentation, and human oversight. European Commission Executive Vice President Henna Virkkunen said the delay aims to “make it easier to innovate without lowering the bar on safety,” citing the need for clearer guidance and technical standards.

Critics argue the postponement leaves vulnerable groups exposed. Annex III of the AI Act classifies certain migration and border management AI as high-risk, but the delayed safeguards mean those systems will not face additional protections for another 16 months. Stefi Richani, advocacy lead at the Equinox Initiative for Racial Justice, said the Act “already undermines the EU Charter’s non-discrimination clauses” and that the delay “will increase surveillance and discrimination.” Existing laws like GDPR remain in force, but campaigners say they do not address all risks.

The transparency rules may have a global impact, as companies often apply EU standards worldwide—a phenomenon known as the “Brussels effect.” However, the Act’s strongest protections do not extend to EU-funded migration surveillance technology deployed in third countries. The phased rollout began with prohibited practices and AI literacy rules in February 2025, followed by general-purpose AI model obligations in August 2025, and now the transparency requirements. The delayed high-risk obligations are expected to apply from December 2027.