EU AI Act extensions leave transparency duties in place

The EU has extended the implementation timeline for many high-risk AI systems to December 2027 and for certain AI-enabled products to August 2028. But transparency duties under Article 50 have already begun to apply, according to TechRadar contributor Ivana Bartoletti, Wipro’s Global Chief Privacy and AI Governance Officer.

Bartoletti argues that organizations should use the extra time to build lasting AI governance capabilities rather than wait for compliance deadlines. As AI tools spread through businesses, governance is harder to add after the systems are embedded in products, services and processes.

She says organizations can extend existing practices in privacy, security, risk management and accountability to address AI. They should also assess whether models are dependable in their intended contexts, whether people can question their outputs, and whether automated processes disadvantage groups. For purchased AI, she recommends examining data handling, model and service changes, incident response, transparency support and the evidence suppliers can provide.

The article also links AI governance with technology sovereignty. Bartoletti says full control of computing infrastructure and supply chains is beyond most organizations, and frames resilience as a more practical goal: understanding key dependencies and avoiding reliance on a single vendor, cloud provider or model. Open-source models can improve code transparency and control, she notes, but do not remove dependence on external computing capacity.

She advocates considering resilience and data sensitivity in technology decisions from the outset. The source article presents the extended timeline as an opportunity to prepare, while emphasizing that transparency duties are already in force.