How does Finn safeguard my privacy?

bunq

Hey bunqers 🌈

At bunq, your privacy is sacred. As a bank, we deal with our users’ most trusted and private data, which is why we maintain the strictest privacy standards. Finn (Beta) is no exception: as your personal assistant, it will only use your data for your benefit.

In this topic, you can learn more about how specifically Finn (Beta) uses your data and how bunq makes it impossible to jeopardize the sanctity of your private data.

How does Finn use my data?

Finn is designed with a focus on personalized and secure data management. Each data type (transactions, activities, etc) has a separate database where each user's data is stored by us. When you ask Finn a question, it retrieves data from this database in a secure way, ensuring no user’s data gets leaked.

Finn is specifically engineered to access and utilize this individual database, ensuring a customized and private experience for each user. This means that Finn's interactions and assistance are tailored to the specific needs and data of the individual user, enhancing the user experience while maintaining a high level of data security 🔐

Does Finn save my data?

Finn doesn’t train on, memorize, or use your data outside the conversations you’ve had with it. In the tech world, it’s called “amnesic”: every time you start a new conversation, Finn starts each conversation with a clean slate, retrieving only the necessary data to provide you with the best possible support. A temporary overview of your data is created to provide you with an answer. This snapshot is created at the moment you initiate the conversation and won’t be used later for any other purposes.

If you choose to proceed with the same chat you've previously engaged with, a portion of the chat's history will be passed. This allows Finn to continue the conversation.

Further information can be found in our Personal and Business Terms & Conditions.

Where is my data stored?

Your data never leaves our perimeters. We also comply with all applicable regulations when it comes to data protection within Europe (eg. GDPR). Partnering with AWS, ChatGPT from Azure and Google on the latest AI technology, we provide an individual and private model with top security and privacy, meaning that it runs only on our infrastructure in our cloud environment. Additionally, your data is retained for a temporary period of 30 days for the purpose of monitoring any misuse.

No personal data will be used, for purposes other than monitoring misuse, by any third party or individuals.

Does my data go to third parties like ChatGPT or Gemini?

The data related to your question is processed using models in collaboration with our partners AWS, ChatGPT from Azure, and Google. However, it's important to note that these partners only process the data we send to them. They do not permanently store it and will not use it for their own training purposes. Furthermore, all data processing activities are carried out within the European Union and therefore protected by GDPR.

Is there any question Finn can’t answer?

Finn won’t respond to any questions about other users, subjects not related to banking, or any language that promotes violence, self-harm, or unfairness. It’s built to maintain a professional manner at all times.

Is my data shared with others?

Your personal information and data will never be shared with others and will remain private to you.

Finn has been improved to provide better recommendations by accessing the recommended places of your contacts list that are available in your Travel tab. This is the only information from your Travel tab that Finn can access.

Does Finn provide accurate information?

Finn strives to provide you with the most accurate information possible. Finn is in its Beta version, meaning it’s continually improving thanks to human input. By using Finn and leaving a thumbs up or thumbs down to each of Finn’s answers, you help Finn improve.

As any new generative AI technology, Finn can sometimes provide unexpected or inaccurate answers. While the likelihood of errors is low, the potential for diverse and unexpected responses is higher.