Germany Launches AI Security Institute Based on British Model

The German National Security Council has approved the establishment of an AI Security Institute in Germany. This institute, inspired by the British model of the AISI, aims to analyze the capabilities and risks associated with modern artificial intelligence models. It also seeks to strengthen international cooperation and develop common standards with foreign partners to assess the impact of these technologies on cybersecurity in Germany.

The IT industry group Bitkom has welcomed this initiative, emphasizing that the "German AI Security Institute (DE-AISI)" should closely follow the approach of the UK AISI. The latter has conducted extensive preliminary testing on the security risks associated with models such as Anthropic's Mythos series and GPT-5.5. Bitkom stresses the importance of recruiting top-tier technical talent capable of competing with leading providers like Anthropic or OpenAI. To attract these experts, the group recommends competitive salaries, agile structures, secure infrastructure, and strong political support.

Furthermore, the European Union is also seeking access to advanced AI models to assess cybersecurity risks. However, no successful cooperation with companies like Anthropic or OpenAI has been announced yet. According to Bloomberg, the European Union Agency for Cybersecurity (ENISA) is expected to gain access to Mythos, although the agency has not commented on this information despite numerous inquiries.

This situation highlights the EU's current dependence on AI technologies developed in the United States and China, as Europe lacks its own cutting-edge models. The providers of these technologies are often closely tied to their respective governments. Anthropic is said to have sent engineers to the U.S. government to assist in deploying models, including for offensive operations. Both the U.S. and Chinese governments have shown they are willing to strictly regulate these companies to align them with their national interests.