BAS21 | News: How to train AI without violating data protection

How to train AI without violating data protection

Anyone who wants to develop and use artificial intelligence in a company must train it with suitable data. According to a survey of 503 companies with 50 or more employees commissioned by the digital association Bitkom, two-thirds (64 percent) of those who deal with AI state that personal data must be used so that the AI provides usable analysis results. "Anyone working with personal data must meet data protection requirements," says Dr Nabil Alsabah, Head of Artificial Intelligence at Bitkom. "Developers have two important tools at their disposal to disguise or completely remove the reference to persons: they can pseudonymise or anonymise the data."  The digital association Bitkom shows how this can be done in practice in a new guide "Anonymisation and Pseudonymisation of Data for Machine Learning Projects".

After an introduction to the various methods of anonymisation and pseudonymisation, concrete applications and measures are presented on how personal data can be used in a data protection-compliant manner and where possible dangers lurk. Among other things, solutions for dealing with geo-movement profiles are presented, as well as the methods Google uses to collect mobility data for the Covid 19 alert. Separate chapters are devoted to the anonymisation and pseudonymisation of medical text data, the process of semantic anonymisation, and data protection risks in media content. In addition, the concept of federated learning is presented in detail, in which machine learning only takes place locally and thus no personal data is transferred to third parties.

Want to learn more about Big Data and AI and how it can be implemented into business processes? On 21 and 22 April 2021 you will have the opportunity to digitally network with AI pioneers, decision-makers, and solution architects. #BAS21 will help you discover novel approaches to creating business value from AI. You can take a deep-dive into real-world industrial use-cases. Over the course of two days, we will also address the ethical and societal challenges of human machine interactions. In short: #BAS21 will take you beyond the hype right into the real world of AI applications. Get your free ticket here: