Privacy-friendly AI for the Public Sector

Press Release

Jun 22

Berlin-based search engine Xayn and Digital Society Institute at the ESMT Berlin publish joint position paper

Governmental AI strategies can be future-proofed with new GDPR-compliant and efficient technology

The public sector can benefit from decentralized AI (Masked Federated Learning)
New position paper explains potential use cases for public health and law enforcement

BERLIN, 22 June 2021 – To implement and future-proof governmental AI strategies, we need powerful and GDPR-compliant technology for the public sector. So-called Masked Federated Learning can meet these high requirements for data protection, trustworthiness, and performance. In a position paper published today, the Berlin-based tech company Xayn and the Digital Society Institute (DSI) at the European School of Management and Technology (ESMT) highlight the potential of this decentralized new approach and elaborate on potential use cases in the health sector and law enforcement.

"Countries that want to play a major role in the global AI competition of the future need efficient and, above all, privacy-friendly technologies for the public sector. New decentralized approaches such as masked federated learning combine both – and can thus become an EU model for responsible AI use," explains Professor Michael Huth, Chief Research Officer at Xayn, Dean at Imperial College London, and co-author of the statement.

"Artificial intelligence also has great application potential in sensitive areas of public services, such as healthcare or public security. Realizing this potential requires privacy-friendly technologies, and they are already available," emphasizes Martin Schallbruch, Director of the Digital Society Institute at ESMT Berlin and co-author of the statement.

Key points of the joint position paper

Masked Federated Learning is a form of distributed machine learning that unites data protection and Artificial Intelligence. The raw data always remains on users’ end devices and trains local AI models. These models are aggregated in encrypted form into a global model and fed back. This approach ensures that data protection is maintained, while at the same time efficiently training GDPR-compliant AI.
The public health sector can benefit significantly from the use of federated learning – for example, in preventing and combating future pandemics. For this purpose, cross-device applications (e.g., smartphone apps for assessing personal risk) can be combined with cross-silo applications (e.g., databases of health authorities on the current infection situation). With this collaborative, decentralized technology, individual and collective risks could be assessed more quickly, while at the same time, the highly sensitive user data remains on the end devices and is thus always protected.
For public safety and law enforcement, federated learning could be used to support investigations across different police agencies. Police departments could thus benefit from the experiential knowledge of other agencies without sharing personal information with each other. Potential areas of application are cybercrime, such as identity theft, or offenses involving images of sexual violence against children and adolescents.

GDPR-compliant, powerful AI can increase efficiency, strengthen citizens' trust in the authorities, and become an EU model for the responsible use of AI.

Download:

Download the joint position paper "Artificial Intelligence for the public sector: Masked Federated Learning as a new privacy-protecting solution" here.

____________________

About the authors:

Professor Michael Huth is the Co-Founder of Xayn and teaches at Imperial College London. His research focuses on Cybersecurity, Cryptography, as well as security and privacy in Machine Learning.

Martin Schallbruch is the Director of the Digital Society Institute at the ESMT Berlin and teaches at Karlsruher Institute for Technology. He also served as a long-time Director General for Information Technology, Digital Society, and Cyber Security in the German Federal Ministry of the Interior.

Leif-Nissen Lundbæk, Ph.D. is the Co-Founder and CEO of Xayn and specializes in privacy-preserving AI. He studied Mathematics and Software Engineering in Berlin, Heidelberg, and Oxford. He received his Ph.D. at Imperial College London.

Dr Clara Herdeanu is Head of Communications at Xayn. She handled public relations for tech companies such as Mozilla, Stack Overflow, and Alteryx at the international PR agency Ballou and also worked for the ebm-papst Group, the global market leader in ventilation technology.

Lola Attenberger is a researcher at the Digital Society Institute at the ESMT Berlin and focuses on cyber security. She has studied at the Hochschule für Wirtschaft and Recht in Berlin.

‍‍_________________________

About Xayn

Xayn is pioneering next genAI for large organizations. The European startup builds secure, sovereign, and sustainable AI systems that leverage the power of language models at scale. Xayn develops its Retrieval Augmented Generative AI as energy efficient and privacy-protecting AI solution “Made In Germany”.

Born out of research at Oxford University and Imperial College London, the Berlin-based company is committed to European values of transparency, privacy, and sustainability and develops its AI system as open-source technology. Founded in 2017 by Dr Leif-Nissen Lundbæk, Professor Michael Huth and Felix Hahmann, Xayn’s academic vision remains, with a workforce comprised of more than 35% PhDs. The start-up has received investment funding of 19.5 million EURO from Global Brain Corporation, KDDI Open Innovation Fund, Earlybird, and Dominik Schiener.

Downloads