New AI Security Institute Established in Berlin

On September 2, 2026, the AI Security Institute (AISI) was established in Berlin. This new authority aims to evaluate and monitor the safety of artificial intelligence (AI). In light of recent incidents at companies such as Anthropic and OpenAI, the necessity for such an institution has become increasingly clear. The AISI will operate as an independent authority and is tasked with reviewing both public and private AI developments for their safety standards.

The establishment of the institute occurs in a context where the risks of AI applications are coming into sharper focus. Experts warn of potential dangers posed by insecure AI models. The authority will develop various methods for risk assessment of AI models, including algorithm analysis, training data verification, and evaluation of the transparency of AI systems. These measures aim to ensure that AI applications are not only effective but also safe.

A central concern of the AISI is collaboration with existing institutions and companies. The institute plans to form partnerships with universities, research institutions, and industry to develop a comprehensive understanding of the safety requirements for AI. These collaborations are also intended to help establish best practices. The founding of the AISI has been welcomed by various stakeholders from politics and business. The Federal Minister for Digital and Transport, Dr. Volker Wissing, emphasized the importance of safety assessment for AI. He stated that the authority will play a crucial role in strengthening trust in AI technologies while promoting innovation in Germany. The AISI will also provide a platform for the exchange of information and experiences in the field of AI safety. Regular conferences and workshops are planned to give professionals from various fields the opportunity to discuss current developments. These events aim to create a network of experts addressing the challenges of AI safety.

Another aspect of the AISI's work will be the development of guidelines and standards for AI safety. These will apply to both developers and users of AI technologies. The goal is to create uniform requirements that ensure the safety of AI applications. The initial steps of the AISI will focus on analyzing existing AI models, with particular attention to identifying vulnerabilities that have previously led to security incidents.

The authority plans to publish its findings regularly to create transparency and raise awareness of AI safety issues. The establishment of the AISI is part of a broader strategy by the federal government aimed at positioning Germany as a leading location for AI development. The authority will play a key role in raising safety standards in the AI industry, thereby strengthening public trust in these technologies. The first reports on the activities of the AISI are expected in the first quarter of 2027, in which the authority will present its progress and insights.