AI Transparency

Where we use AI

First in translation: Modern language models produce raw translations in up to 24 languages. Second in image production: The illustrations on this website are created with AI image generators based on our specifications and are curated by us. Third in text creation: Drafts for content can be created with AI support. In all three cases, the principle applies: AI delivers drafts, humans are responsible for the results.

Where humans decide

Every core page translation is reviewed by native-speaking proofreaders; terminology decisions are made by humans and documented in the term base. Publications, offers, prices, and all customer communication are approved by humans. At Baduno, there are no fully automated decisions with legal effect on customers within the meaning of Art. 22 GDPR.

Labeling of AI-generated content

Images on this website are AI-generated original productions by Baduno GmbH; we indicate this centrally here and upon request in individual cases. For client projects, we contractually agree on the labeling practice – from silent tool usage to explicit labeling of every asset, depending on the client's industry requirements.

Data & Confidentiality in AI Use

Customer content is transmitted to AI services only to the extent necessary; confidential passages are removed or pseudonymized beforehand, unless expressly agreed otherwise. We do not use customer content to train our own models and select services based on the condition that inputs may not be used for provider training.

Boundaries We Set for Ourselves

No AI creation of legal or medical statements without expert review, no automated evaluations of individuals, no deepfakes or impersonation of real people, no content that infringes third-party rights. Human reviewers examine complaints; for feedback, contact us at [email protected].

Step-by-Step Practical Example

Imagine a client commissions Baduno to translate a product page for the Japanese market. The process begins with the source text being handed over to our system. First, the text is converted by an AI-powered language model into a raw translation for Japanese. This raw translation serves as a draft and is never published directly.

In the next step, a native Japanese reviewer takes control. They compare the raw translation with the original, adjust terminology, check cultural appropriateness, and ensure the client's corporate language is accurately represented. All changes are documented in the central termbase to ensure consistency for future projects.

After linguistic review, the text goes through content approval by a project manager. Quotes, prices, and client-related decisions are made exclusively by humans – no AI is involved here. Only after this approval is the translation delivered to the client. Should the client request images to appear on the page, these are created by an AI image generator according to our specifications but are also curated and approved by humans. This ensures that every step is traceable and controlled.

Common Misconceptions

A common misconception is that AI at Baduno makes autonomous decisions. In reality, the opposite is true: AI merely provides drafts – whether for translations, texts, or images. Every output is reviewed by qualified humans before being used in a client project or on this website. There are no fully automated decisions with legal effect.

Another misconception concerns labeling: that AI-generated content on this website is centrally labeled does not mean we conceal the use of AI in client projects. On the contrary – we contractually agree with each client on how transparently AI usage is to be presented. Some clients want complete labeling of every AI asset, others see AI as a pure tool and forgo explicit marking. Both are possible and are defined individually.

It is also sometimes assumed that client data is used to train AI models. This is not the case: we do not use client content for our own training purposes and select AI services based on the condition that inputs may not be used for the provider's training. Confidential passages are removed or pseudonymized before transmission, unless agreed otherwise.

Internal workflow at Baduno

Internally, the use of AI at Baduno follows a clear, multi-stage process. First, the need is identified: for which task can AI be used supportively without compromising quality and control? Typical applications are the creation of raw translations, text drafts, or image generation. For each application, we define guidelines on which AI tools may be used and how human review must be conducted.

The implementation begins with the selection of the appropriate AI service. We ensure that the service meets the required data protection and confidentiality standards. After an AI draft is created, it is forwarded to the responsible department: translations to native-speaking reviewers, texts to editors, images to the creative team. The reviewers document changes and only release the draft once it meets the agreed quality criteria.

Finally, quality assurance is carried out by the project manager, who also checks compliance with the labeling agreements with the client. All steps are documented so that it is always traceable whether and how AI was involved. Responsibility for the final result always lies with humans – never with the machine.

What this means for clients in concrete terms

For you as a customer, the transparent use of AI at Baduno means one thing above all: you retain control over your content and its configuration. Concretely, this means that for every project processed with AI support, you can request information about which AI tools were used and to what extent. You can at any time commission purely human processing – naturally with a corresponding adjustment of the scope of services and costs. Furthermore, you have the option to define the labeling of AI-generated content according to your own requirements. Our standard is the central labeling on this site; however, if your company wishes individual labeling of each asset, we agree this contractually. Moreover, you can be assured that your data will not be used for training AI models and that confidential content will be pseudonymized or removed before transmission to AI services, unless otherwise agreed. Should you have any questions or complaints, our team is available at any time – your contact person will personally review your concern and not delegate it to an AI.

Terms explained clearly

To help you better understand the processes at Baduno, we explain some key terms here. By raw translation we mean the initial machine transfer of a source text into a target language. It serves as a draft and is never used unchecked. A term base is a central database in which preferred terms and translations are recorded – this ensures consistency of terminology across all projects. An AI image generator is software that creates independent illustrations based on text prompts. At our company, these prompts are formulated by humans and the results are curated by humans before being used. A native-speaking reviewer is a language specialist whose native language is the target language. He or she compares the raw translation with the original, corrects errors, checks cultural appropriateness, and ensures that the customer's corporate language is correctly reflected. The term approval describes the final step in which a project manager reviews the finished work and authorizes it for delivery – only humans make this decision. Pseudonymization means replacing personal data with artificial identifiers so that the content can no longer be directly attributed to a person.

Interaction with our other guidelines

This AI transparency declaration is consistent with other policies of Baduno GmbH. In particular, it supplements our privacy policy, in which we describe how we handle personal data. When using AI, we ensure that personal data is not processed without control. Customer data that passes through our systems is subject to the same confidentiality and deletion periods as all other order data.

Our quality assurance policy also applies here: every AI-generated draft undergoes a human review process that follows the same quality standards as purely human work. Documentation in the termbase is performed according to the same rules. Should a customer require specific labeling obligations – for example, for regulated industries such as medicine or finance – we agree on these contractually.

Furthermore, we refer to our General Terms and Conditions, which govern liability for AI-generated content. We make it clear that we are only responsible for reviewed and approved content. The AI Transparency Declaration itself is not a legally binding part of the contract, but serves to inform and build trust. It is updated regularly to reflect the state of the art and legal requirements.

What this means for clients

For clients, our use of AI primarily means efficiency and consistent quality. They benefit from faster draft translations that our experts then refine. AI serves as a tool, not a decision-maker. Therefore, clients retain control at all times: they never receive automatically generated end products, but always results that have been reviewed and approved by humans.

If you commission a website localization from us, you receive transparent information about where AI has provided support. Upon request, we explicitly label individual assets. For projects with special confidentiality requirements – such as intellectual property or content not yet published – we adapt the use of AI services. We eliminate sensitive passages before processing or rely on locally running models.

Moreover, you can be confident that your content will not be used to train AI models. We select our service providers specifically based on the criterion of non-further use. And if you have questions about a specific translation or illustration, we will disclose the exact creation process. In short: our use of AI changes the workflow, but not your rights or our duty of care.

Internal process at our company

Before an AI-powered tool is deployed at Baduno, it undergoes an internal review. Our team evaluates whether the service meets the requirements for data protection, confidentiality, and quality. We pay particular attention to ensuring that input data may not be used by the provider for training their own models.

The operational process begins with the intake of a client project. Our system captures the requirements and assigns the appropriate tools – whether it be a language model for translations or an image generator. The AI produces a rough draft, which is automatically transferred to a review environment. There, native-speaker reviewers and subject matter experts work on refinement. All changes are documented in our termbase to ensure consistency across projects.

Only after final approval by a project manager does the result leave our company. No AI systems are involved in this step – the decision lies solely with humans. In this way, we ensure that each step remains traceable and that control rests with humans. Internal audits and random checks complement the process to verify compliance with our standards.