Thulium AI and personal data protection

As part of Thulium AI, we enable you to use three new features: AI Assistant, AI Translations, and AI Voice Analytics.
You can find out more about how you can use the new features of the Thulium System here:

In this article, you will learn how we take care of the processing of your personal data when using Thulium AI.

AI assistant

The AI assistant is built into the core of the product, and there are no additional fees for using it. The AI assistant is based on GPT models running on Microsoft Azure infrastructure, which we use to process and store your data.
The AI assistant has been configured to ensure your organization's compliance with the GDPR, including data processing within the EEA. This means, among other things, that your input and output data:

They are NOT available to other customers,
They are NOT available to OpenAI,
They are NOT used to improve OpenAI models,
They are NOT used to train, retrain, or improve the Azure OpenAI service,
are NOT used to improve Microsoft or third-party products or services without your consent or instruction,
a given AI assistant is available exclusively to a specific customer, which means that the data processed using it remains completely separate from the data of other customers.

The Azure OpenAI service implemented in the Thulium System is supported by Microsoft as an Azure service. Microsoft hosts OpenAI models in the Microsoft Azure environment, and the service does NOT work with any services supported by OpenAI (e.g., ChatGPT or the OpenAI API).

Using the AI assistant means processing data within the EEA.
You can find more information on this subject here.

AI Translations

AI Translations is a paid add-on for translations, without the possibility of prior testing.

The functionalities of AI Translations are provided using AI Assistant mechanisms, i.e. based on Azure OpenAI, with full application of the data processing principles described above, and additionally using DeepL functionalities.

In the context of DeepL, we use the DeepL Pro version, which offers increased confidentiality protection. With this version, DeepL only temporarily stores both the original content and the processed results, and only to the extent technically necessary to provide the services, and then deletes them. The data is never used to train artificial intelligence models.

The use of AI Translations means that data is processed within the EEA.

AI Voice Analitycs

AI Voice Analytics is a paid add-on used for transcription and basic analysis of conversations, with limited free use.

The functionalities of AI Voice Analytics are provided using AI Assistant mechanisms, i.e. based on Azure OpenAI, with full application of the data processing principles described above, and using the functionality of AssemblyAI.

The use of AI Voice Analytics involves data processing within the EEA, subject to further clarification. The use of AI Voice Analytics means that your organization's data will be processed using so-called endpoints within the EEA. This means that processing takes place primarily in infrastructure located in Europe. We rely on AssemblyAI technology, which is a US-registered entity. This means that data will be stored and processed within the EEA, and access from the US will only be possible in limited cases, e.g. for technical support. Such potential data transfers are secured by legal mechanisms recognized by the European Union (standard contractual clauses and additional safeguards agreed between us and AssemblyAI, such as encryption, MFA, VPN, managed, limited, and supervised access).

Remember also that using AI Voice Analytics allows you to understand the sentiment of the caller, i.e. the emotional attitude expressed during the conversation. The sentiment of the caller is based on the phrases used by the caller. Sentiment may refer to a specific caller or constitute aggregated statistical information. The use of AI Voice Analytics may therefore mean that your organization is using profiling, which should be subject to your individual assessment. Therefore, make sure that you have the appropriate legal basis for processing data in this way, that you fulfill your information obligation, and that you comply with other legal obligations arising from this fact.

In summary, when implementing all Thulium AI functionalities, we have applied a number of security measures to ensure a high level of protection for your data. In this context, the following mechanisms should be mentioned in particular:

Separation of customer data
Data encryption, restricted and supervised access, MFA, VPN
Data processing within the EEA (with exceptions regarding AssemblyAI)
No processing of data for model training, machine learning, or model improvement
Operation based on appropriate entrustment agreements

Please note that by using Thulium AI, you accept the above data processing rules.