Introducing the future of localization

Agentic AI takes action, not just conversation. Athena uses goal-oriented agents to help you complete real tasks like managing vendors or spotting risks — not just answering questions.

Yes. You can integrate your own licensed LLMs and host them privately, giving you full data control while benefiting from real-time context injection.

SmartContext and Language Guard support 90+ languages, with ongoing expansion based on customer needs.

Yes. Intelligent Score is designed to work across human, machine, and LLM-generated translations, providing consistent quality evaluations no matter the source.

XTM GenAI features are embedded across the entire localization lifecycle to deliver context-aware translations, assess and improve quality and optimize project management with context-aware decision support.

XTM GenAI features leverage third-party hosted models to help your team work smarter from start to finish.

To learn more about each feature’s use case and functionality, refer to our Features pages:

• SmartContext
• Language Guard
• AI Translation and Intelligent Score
• xaia

All AI features in XTM are optional. Users can flexibly decide to turn these features on. By default all AI features are turned off. We also support the concept of bringing your own key. It means that you have the flexibility of plugging your own LLM key (provided that the LLM is supported in the product) and process your content through your own LLM. Below is a list of GenAI features and LLM options:

• Language Guard: it runs on your own OpenAI GPT subscription. This means that the feature can be enabled only if you bring your own key.
• AI Translation and Intelligent Score: you can leverage the XTM-managed LLMs or bring your own key. This feature leverages OpenAI GPT and Gemini models.
• xaia: this feature is based on Gemini models. It is not possible to bring your key for this feature but we can give you the option to own and manage the relevant Google Cloud Project. Refer to our Responsible AI principles document for more information.

Customer data is not being retained or used to improve or train the LLM(s) or the XTM products and services. LLM training is disabled at API level. All data is encrypted in transit (TLS) and at REST (AES-256).

Users can flexibly turn AI features on or off. When turned on, a feature needs to rely on the LLM it leverages. It is not possible to deactivate a specific LLM for a feature, as this would impact its correct functionality.

Yes. When GenAI Features are enabled, data processed through these features may be transferred to third-party LLM providers, such as OpenAI or Google, depending on the specific feature in use. As such, data residency is determined by the infrastructure and regional hosting policies of these LLM providers. We ensure that all data transmitted to LLMs is encrypted in transit (TLS 1.2 or higher), and that LLM training is explicitly disabled at the API level to prevent data retention or reuse. Customers may consider using the “bring your own key” option, were supported, to maintain more control over data processing and residency.

In line with our Transparency principle, all AI features are optional and turned off by default, so you can decide if and to which extent to use these functionalities.

Yes, our Subscription Agreement Terms and AI solutions addendum shall apply to GenAI Features found on our security and trust centre.

Once enabled, all users on your instance will have access to the GenAI features, in line with their user roles and access rights. Each functionality might show different options depending on the user role. For more information, refer to our user documentation for XTM and user documentation for XTRF.

XTM GenAI features are designed with strict data protection controls. All data is encrypted in transit and at rest, and is never used to train language models — training is disabled at the API level. Access is controlled through your existing user roles and permissions, and each feature uses only the data required for its function.

Features like xaia access project data in real time via secure APIs without storing or learning from it. All GenAI features are optional and off by default, with the option to bring your own LLM key where supported. XTM is ISO 27001 certified and has a formally defined and secure SDLC to ensure enterprise-grade security.

Data residency for GenAI features depends on the underlying infrastructure of the third-party LLM providers, such as Google and OpenAI. While most features rely on standard cloud hosting by these providers, certain features offer more control over data location.

For example, some capabilities can be deployed under your own cloud project (e.g., Google Cloud for Gemini-based features) or routed through region-specific endpoints (e.g., Azure-hosted OpenAI), allowing you to align data processing with your preferred geographic region.

Although not all features support custom residency at this stage, XTM is committed to enabling greater flexibility where technically feasible, while ensuring all data is encrypted, transient, and never used for model training.

As a company, we are fully compliant with GDPR as well as with new European regulations such as the AI Act. These two acts complement each other.

To control compliance, we have created a cross-team team to ensure that our GenAI solutions comply with applicable laws and regulations.

Each GenAI feature relies on specific data to guarantee its correct functionality. Once enabled, it is not possible to influence or modify the data each feature relies on.

• Language Guard uses your source content and translation to identify inappropriate language and suggest a more appropriate alternative in line with the source text
• AI Translation and Intelligent Score use your source content, translations, TM fuzzy matches, terminology and subject matter to provide a context-aware translation and a confidence score to assess its accuracy.
• xaia is not pretrained or trained on your data. Instead, it retrieves relevant project information in real time using a “tools-based” approach — meaning it interacts with XTM APIs at the moment a user submits a question. No data is stored or retained, and no user data is used to train the model.

Yes, all GenAI features fully adhere to the role-based access control (RBAC). Each feature operates strictly within the access rights of the user invoking it, ensuring that users can only access and process data they are already authorized to see. This applies consistently across all GenAI functionalities and aligns with your existing permission model.