The EU AI Act entered its general application phase on August 2, 2026, but this does not mean every AI system takes on the same obligations on the same day. The regulation applies in phases based on use-case risk, whether a system is embedded in a regulated product, and the company’s role in the value chain. Multinationals should not determine their compliance scope simply by asking “Do we use AI or not?”
The European Commission’s current timeline shows: rules on prohibited AI practices and AI literacy have applied since February 2, 2025; governance and obligations for general-purpose AI models have applied since August 2, 2025; and certain transparency requirements apply from August 2, 2026. After the AI Omnibus entered into force on July 27, 2026, high-risk uses listed in Annex III were postponed to December 2, 2027, while Annex I high-risk systems embedded in regulated products such as machinery, toys, or lifts were postponed to August 2, 2028.
The first step, therefore, is not to buy a “compliance tool” right away, but to build an AI system inventory. Each entry should record the purpose of use, countries of deployment, model and system vendors, whether the company may act as provider or deployer, input and output data, affected persons, whether AI interactions are disclosed externally, and whether the system touches employment, education, critical infrastructure, or other potentially high-risk uses. Final classification and obligations must still be determined by legal or compliance professionals based on actual processes and target jurisdictions.
The second step is to connect controls to operational processes: assign system and data owners, retain vendor documentation and versions, define human review points, log incidents and complaints, restrict permissions for high-impact actions, and set deactivation and recovery procedures. If the model, use case, data source, or deployment region changes, the system should be reassessed rather than relying on the conclusions reached at initial rollout.
The third step is to verify in batches according to the timeline. For the prohibition, literacy, general-purpose AI, and transparency requirements already in application, complete a gap assessment first. For systems that may fall under Annex III or Annex I high-risk categories, use the transition period to assemble evidence on risk management, data governance, technical documentation, record-keeping, human oversight, robustness, and cybersecurity. A postponed application date does not mean governance can be paused; it gives companies time to build auditable readiness records for clearly defined use cases.
From the perspective of “The CIO for Your Business Operations,” Unima recommends managing cross-border AI adoption with five fields: use case and market, company role, data and vendors, human accountability, and evidence and timeline. This is an operational review framework compiled from publicly available European Commission information and does not constitute legal advice, a regulatory applicability conclusion, certification, or a compliance guarantee.
First-party sources: European Commission, “AI Act,” https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai; “AI Omnibus enters into force,” 2026-07-27, https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force; “The enforcement framework of the AI Act,” https://digital-strategy.ec.europa.eu/en/policies/enforcement-ai-act. This article was re-verified on 2026-08-26.