paluno: Zoltán, the session “Privacy-preserving technologies – a key enabler of big data for AI” had approx. 100 participants. So the interest was great. What was the specific topic of the session?

Zoltán Mann: : In the session, we focused on the technological possibilities to bring AI into line with the requirements of data protection. The EU is supporting several research and innovation projects that address the issue from different perspectives. Interesting results from some ongoing and recently completed projects have been presented, e. g. from the projects Papaya, Kraken, MOSAICrOWN und Musketeer. 

The objective of the session was to enable researchers to network across the projects in order to discuss possible technical solutions for data protection.

paluno: What technical challenges arise when artificial intelligence meets data protection?

Zoltán Mann: The challenges are diverse. For example, the security of the data is to be ensured in the sense of security and in addition to protecting the privacy of the users. Generally, different security and privacy protection approaches must be combined to meet all data protection requirements. The use of artificial intelligence makes it more difficult to predict the system’s behaviour when designing the system, even though data protection should be ensured “by design”.

The approaches presented at the EMCDDA are promising, but in some cases they are not yet fully developed. Many are still afflicted with certain “side effects”, such as: a negative impact on performance or user-friendliness of the applications. So there’s still a lot to do in the field.

paluno: What new questions do AI applications raise?

Zoltán Mann: Many AI applications are based on machine learning. Such applications only work if models for the given purpose are trained on the basis of realistic data. The training data may also include sensitive personal data. Can someone draw conclusions on the personal data of individuals on the basis of the models? What protection mechanisms need to be put in place here? How can we guarantee that the training data remains protected? These are questions that need to be clarified in order to prevent breaches of data protection.

paluno: Can AI also help to better protect data?

Zoltán Mann: Yes, AI could, for example, help to detect attacks by hackers on IT systems. It could also help to automate defence or protection mechanisms in the future. But this is only one side of the coin. Hackers can also use AI to perform for example to spy on the behaviour of users or to deceive protective mechanisms.

paluno: So artificial intelligence is a powerful challenge for data protection. Thank you for the insights into this important field of research.