Data protection with AI: Experts call for legal certainty for the use of technical solutions
Munich, 09 October 2023
Artificial Intelligence (AI) can make an important contribution to a sustainable economy and society – be it through optimized business processes or products that promise personalized healthcare, safer road traffic or better educational opportunities. At present, however, companies are still reluctant to use AI. The frequently cited reason is that the data protection hurdles appear to be too high. A current white paper from Plattform Lernende Systeme shows technical approaches that ensure privacy and data protection in the development and use of AI systems. The authors call for legal recognition of the procedures in order to strengthen legal certainty for companies using AI.
Data protection enjoys a high priority in Germany and Europe. At the same time, companies often hold valuable data that could be harnessed for the benefit of society with the help of AI technology: For example, patient health data could be used to better predict the development of diseases; movement data of people and vehicles could be used to reduce risks in road traffic.
Legislators impose strict requirements on any use of personal data – with legal interpretation often uncertain in practice, making AI use more difficult across the board, according to the whitepaper “Harnessing data treasure for AI, preserving data privacy with AI.” For this reason, many companies are reluctant to deploy and develop AI systems that process sensitive user data. However, various technical privacy approaches exist that make it possible to maintain data privacy when using personal data. The authors of the white paper therefore call for technical instruments for more privacy to be legally recognized. The procedures should be included as an exception in the European Union’s General Data Protection Regulation (GDPR) and future AI Regulation and formulated in application-specific data protection legislation. This would allow for a more flexible use of personal data. Prerequisite: the use of personal data has no alternative and is in the interest of the common good.
Creating scope for action for AI development
“The use and development of AI require legal certainty. Instead of prohibitions, legislators should create room for maneuver and legally permit technical procedures to ensure data protection. In this way, currently existing room for interpretation in the processing of personal data can be closed and the opportunities of the key technology AI for our society can be better exploited,” says Jörn Müller-Quade, professor of cryptography and security at the Karlsruhe Institute of Technology (KIT) and co-leader of theIT Security, Privacy, Legal and Ethical Framework working group of Plattform Lernende Systeme.
In concrete terms, the experts recommend, for example, privacy-preserving machine learning (PPML for short), which ensures data protection as early as the design stage of the AI application. This includes the anonymization, pseudonymization or encryption of personal data. They also mention technical approaches that do not start directly with the AI model, such as the use of personal information management systems (PIMS) or data trustees, with the help of which data providers can retain control over their data and even profit from its monetization. Explainable AI, i.e. AI systems that make their decisions and functioning transparent and understandable, can further strengthen self-determined handling of one’s own data. Standards and certification options should be introduced for approaches to explainable AI as well as for anonymization of data.
The authors of the white paper emphasize that for the training of AI systems, non-personal data should generally be preferred to personal data, provided they have the same data quality. They therefore recommend building interoperable data spaces to make more non-personal data available.
Dr. Detlef Houdeau, co-author of this white paper, also provides initial answers on the topic of “Data treasure and data protection” in the “3 questions to” video series (in German) of Plattform Lernende Systeme.
About the white paper
The white paper “Harnessing Data Treasure for AI, Preserving Privacy with AI. Technical and legal approaches for privacy-compliant, public good-oriented data use” was written by members of the IT Security, Privacy, Legal and Ethical Framework working group of Plattform Lernende Systeme. It is available for free download (in German).