Provider or Deployer: Who’s Responsible for AI Translation Under the EU AI Act? 

As businesses become more reliant on AI-powered tools, an important question emerges: who’s responsible when something goes wrong?

Imagine an AI-assisted translation introduces an error into a product safety warning, or an inaccurate translation leads to a contractual dispute. Is responsibility shared by the company that created the AI system, the language service company (LSC) that used it, or the organisation that commissioned the translation?

Under the EU AI Act, the answer depends on the role each organisation plays. As such, understanding the distinction between providers and deployers is crucial for businesses using AI-assisted translation, as it influences transparency, governance and compliance obligations throughout the translation process.

In this article, we’ll explain what the differences between providers and deployers mean in practice, explore how these distinctions apply to AI-assisted translation, and highlight the questions translation buyers should ask when selecting an AI-enabled language company.

Provider vs. Deployer: What the EU AI Act Actually Says

The EU AI Act distinguishes between two key roles: the provider, which develops or places an AI system on the market, and the deployer, which uses that system in a professional context. Understanding the difference is important because each role carries its own regulatory obligations, from transparency and documentation to human oversight and risk management.

In simple terms:

• Provider: The organisation that develops an AI system or places it on the market under its own name. In AI-assisted translation, this could be the company behind a machine translation engine or another AI-powered language tool.
• Deployer: Any organisation that uses an AI system in the course of its professional activities. Depending on how AI is used, this may be an LSC, an organisation using AI translation internally, or another business integrating AI into its translation workflows.

While providers and deployers have different responsibilities under the AI Act, one important point is often overlooked: using a third-party AI tool does not automatically transfer responsibility to its developer. Organisations deploying AI still have obligations of their own, particularly around how the technology is used, monitored and integrated into professional workflows.

For translation buyers, this means that evaluating an LSC’s AI practices involves more than asking which tools they use. It’s just as important to understand how those tools are governed, what level of human oversight is applied, and how quality and compliance are maintained throughout the process.

When Does AI Translation Require Greater Oversight?

The amount of responsibility providers and deployers have isn’t always the same. Much depends on the type of content being translated and the potential consequences of an error.

For example, while a misprint in a promotional email may be embarrassing, a mistake in a medical leaflet, legal contract or product safety document could have far more serious consequences.

The EU AI Act recognises that AI systems used in certain regulated sectors, or for particularly sensitive purposes, may require greater oversight than those used for everyday activities. Consequently, the obligations placed on organisations depend not only on the AI system itself, but also on how and where it’s used.

For most organisations, AI-assisted translation of routine business content is unlikely to fall within the AI Act’s high-risk categories. However, businesses operating in regulated industries or translating content with legal, medical or safety implications should take additional care to ensure AI is being used responsibly, with appropriate human oversight and quality assurance.

The practical takeaway is that the greater the potential impact of a translation error, the greater the need for transparency, governance and human review. Rather than asking whether AI can be used, translation buyers should consider whether the safeguards in place are appropriate for what’s being translated.

Three Practical Questions Translation Buyers Should Ask

Understanding the distinction between providers and deployers is only part of the picture. Translation buyers must also understand how AI is used within their LSC’s workflows, particularly for projects that rely on accuracy, compliance and confidentiality. By asking the right questions, buyers can clarify how responsibility is managed in practice.

1. Who configured the MT system?

The way an AI system is implemented can affect who is responsible for how it’s used. If an LSC has customised a machine translation (MT) engine through fine-tuning, domain adaptation or integrating its own glossary, it may assume additional responsibilities beyond simply using an off-the-shelf tool.

For translation buyers, this reinforces the importance of understanding not only which AI tools an LSC uses, but also how they have been configured and integrated into the translation workflow.

2. Is there documented human oversight?

Human oversight remains one of the key principles of the EU AI Act. While AI can improve efficiency, organisations must ensure appropriate review processes are in place, particularly for content that could have legal, medical or commercial consequences.

For many LSCs, this means using Machine Translation Post-Editing (MTPE), where qualified linguists review and refine AI-generated translations before delivery.

3. What does the Data Processing Agreement say about AI?

Many organisations already have Data Processing Agreements (DPAs) in place with their language partners to govern the handling of personal data. However, these agreements may not address how AI is used, who is responsible for AI-generated outputs or how incidental personal data is handled when AI-assisted translation forms part of the workflow. Translation buyers should therefore look beyond data protection alone and consider whether AI-specific responsibilities, transparency measures and governance arrangements are clearly documented.

A responsible LSC should be able to explain not only which AI tools it uses, but also the policies, processes and human oversight that help ensure quality, transparency and regulatory compliance. This means having documented AI governance, secure translation workflows, contractual safeguards and clear quality assurance procedures that support both the EU AI Act and broader data protection obligations. 

Five Questions to Ask Your LSC

As AI-assisted translation becomes the norm, asking the right questions can help organisations identify language service companies that take AI governance seriously. While every organisation’s requirements will differ, the following questions provide a useful starting point when evaluating an LSC’s approach to AI-assisted translation:

1. Do you have a documented AI usage policy for your translators and product managers?
2. Is human post-editing mandatory, or is raw MT output ever delivered?
3. Which MT engines do you use, and how have you assessed them for compliance and suitability?
4. Does your DPA include clauses specific to AI-generated content and liability?
5. Can you confirm that your workflows comply with the EU AI Act deployer obligations?

A language service provider should be able to answer these questions clearly and explain the governance, quality assurance and human oversight that underpin its AI-assisted translation workflows. By choosing an LSC that combines robust AI governance with experienced human linguists, organisations can benefit from AI’s efficiency while maintaining the quality, transparency and compliance their multilingual content demands.

If you’re looking for an AI-assisted translation partner, Word Connection combines advanced language technology with experienced human linguists to help organisations deliver high-quality, compliant multilingual content. 

Get in touch to discuss how we can support your translation needs.