First EU AI Act Provisions Apply: Compliance Moves from Paper to Practice

On February 2, 2025, the first provisions of the EU Artificial Intelligence Act began to apply: the list of banned "unacceptable-risk" AI practices took effect, and the "AI literacy" article now requires organizations providing or using AI systems to ensure their personnel have sufficient AI knowledge and skills. The world's first comprehensive AI law has moved from paper into practice.

The first batch of provisions has limited direct impact on language services — machine translation and AI writing are not on the banned list. But the "AI literacy" clause carries deep meaning: enterprises providing or using AI translation services in the EU market need to show their staff understand the capabilities and limits of AI systems. This aligns exactly with the industry's capability-building direction of the past two years — training linguists to command AI tools now has statutory backing.

Over the coming year, the Act's obligations land in waves: in August 2025, transparency duties for general-purpose AI model providers begin to apply; in August 2026, the bulk of provisions apply in full. LSPs serving the EU market need to build the compliance timeline into service design.

For enterprise clients, choosing a language provider gains a practical consideration: whether the provider's AI use is disciplined and traceable — no longer a bonus, but a link in the EU compliance chain.