Germany's National Security Council greenights an AI Safety Institute modeled after the UK's AISI
Quick Answer
Germany's National Security Council has approved the establishment of the DE-AISI, an AI safety institute modeled after the UK's AISI.
Quick Take
This institute will assess security risks of frontier models from companies like Anthropic and OpenAI, highlighting Europe's reliance on US and Chinese AI technologies until it develops its own models.
Key Points
- DE-AISI will evaluate AI models for security risks.
- The institute follows the UK's AISI framework.
- Europe currently lacks its own frontier AI models.
- Germany's initiative highlights dependency on US and Chinese technologies.
- The move aims to enhance AI safety in Europe.
📖 Reader Mode
~2 min readThe National Security Council, chaired by the Federal Chancellor, has decided to set up an AI Security Institute in Germany.
The institute will analyze the capabilities and risks of modern AI models, boost cooperation with similar bodies abroad, and develop shared standards with international partners. The goal is to assess how advanced AI models affect cybersecurity in Germany.
According to IT industry group Bitkom, the "German AI Security Institute (DE-AISI)" is explicitly modeled after Britain's UK AISI. That organization has drawn international attention with extensive pre-release testing on security risks in Anthropic's Mythos series and GPT-5.5. "A DE-AISI should follow the British model of the UK AI Security Institute and its unique approach as closely as possible," Bitkom writes.
According to Bitkom, DE-AISI needs top technical talent with an international reputation to test AI models on equal footing with frontier providers like Anthropic or OpenAI. The group is pushing for salaries outside standard public pay scales, agile structures, secure infrastructure, and political backing at the highest level.
Germany and the EU depend on China and the US for frontier AI
The EU is also trying to get access to frontier models to test them for cybersecurity risks. So far, though, there's been no official word on any successful cooperation with Anthropic or OpenAI. According to Bloomberg, ENISA, the European Cybersecurity Agency, is set to receive access to Mythos. Even after repeated requests over several days, the agency couldn't comment on the report.
The situation shows just how dependent EU countries are on US and Chinese AI technology as long as Europe doesn't build its own frontier models. The model providers are tightly linked to their home governments. Anthropic reportedly sends engineers to the US government to help deploy models, including for offensive operations. Both governments have also shown they're willing to heavily regulate AI companies and tie them to state interests.
— Originally published at the-decoder.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.
