As a post-editing agency in Switzerland, it is supremely important that we ensure data security when using machine translation technologies such as DeepL. But even beyond DeepL, we ensure optimal protection of your data at all times. Here are some of the key points we consider:
Encryption of data
All translation data that we send to DeepL is secured before transmission using modern encryption standards such as AES-256. This ensures that confidential content remains protected during transmission to DeepL. In addition, we use a secure connection (HTTPS) for data exchange with DeepL to prevent man-in-the-middle attacks.
Federated learning
DeepL uses a technology called ‘federated learning’, where the translation models are trained on the users’ servers. This means that no texts or translation data need to leave DeepL servers. Only the anonymised model updates are transferred, increasing data security. This process means that the original texts always remain under the user’s control and are not passed on to third parties.
Differential data protection
DeepL also uses what is known as differential data protection to prevent translations from being traced back to individuals. This means that even if data were intercepted, a specific user could not be identified from it. The use of techniques such as differential privacy and k-anonymity ensures that the translation data cannot be linked to specific individuals.
Monitoring and logging
We carefully monitor all access to DeepL and log it. This allows us to quickly recognise suspicious activity and respond to possible data protection breaches. The logs are stored in encrypted form and regularly checked for anomalies. If necessary, we can also make the logs available to supervisory authorities or law enforcement agencies.
Compliance with data protection guidelines
It goes without saying that we comply with all applicable data protection guidelines and laws when using DeepL, in particular the Swiss Data Protection Ordinance (DPO) and the European General Data Protection Regulation (GDPR). Our top priority is therefore to give our customers the assurance that their data is secure. We have adopted internal guidelines and procedures to ensure compliance with data protection regulations and regularly train our employees on data protection issues.
Transparency and control
We attach huge importance to transparency and give our customers full control over their data. On request, customers can see which of their data has been transmitted to DeepL and can have it deleted at any time. We also inform our customers about changes at DeepL or in our own data protection practices.
Non-disclosure agreements
Individual agreements on the strict confidentiality of texts are of course possible at any time. We regularly conclude such agreements with our customers, especially when strict confidentiality is required. No information originating from the customer’s texts is passed on to third parties – and every employee working on these texts is obliged to maintain absolute confidentiality.
As a post-editing agency in Switzerland, these are the comprehensive measures we take to ensure data security when using DeepL. Our customers can rest assured that their confidential translation projects enjoy the best possible protection. We take the issue of data protection very seriously and are constantly working to improve our security measures and keep them up to date.
References:
https://azure.microsoft.com/de-de/resources/cloud-computing-dictionary/what-is-deep-learning
https://www.alexanderthamm.com/de/blog/deep-learning-in-der-praxis
https://de.linkedin.com/advice/1/how-can-you-secure-data-when-using-deep?lang=de
