Bitkom calls for a high-performance AI Security Institute modelled on the British example

In Germany, the debate on artificial intelligence is often dominated by innovation funding, regulation and economic potential. However, as the capabilities of so-called frontier AI models continue to grow, another aspect is coming increasingly to the fore: the national security dimension of AI.

The digital industry association Bitkom is therefore calling for the establishment of an independent German AI Security Institute (DE-AISI). The aim is to enable the German government to better understand the capabilities of modern AI systems, identify risks at an early stage and make security policy decisions based on a robust technological foundation.

AI is becoming a factor in national security

Modern AI models are developing capabilities that go far beyond traditional automation tasks. They are already supporting software development, data analysis and cybersecurity processes. At the same time, new risks are emerging from automated cyberattacks, disinformation, manipulation or the misuse of powerful models for security-critical purposes.

According to Bitkom’s assessment, Germany currently lacks a comprehensive overview of the actual capabilities of such systems and their impact on the state, the economy and critical infrastructure.

In particular, the latest advances by international AI providers demonstrate how rapidly technological capabilities can develop. Security authorities and policy-makers are thus faced with the challenge of having to assess developments that often progress faster than traditional political processes.

Not a regulatory authority, but a technology centre

Unlike existing supervisory or regulatory bodies, the proposed institute is not intended to be a supervisory authority.

Rather, the call is for a technically oriented centre of excellence that focuses on the analysis and evaluation of advanced AI systems. The focus is on technical testing, risk assessments and the scientific investigation of potential impacts on Germany’s security and digital sovereignty.

In particular, the institute is intended to provide insights into

• the capabilities of new AI models,
• the potential for misuse,
• which protective measures are effective,
• and the strategic implications for the state, the economy and critical infrastructure.

Focus on cyberattacks and critical risks

Particular attention is to be paid to security-relevant fields of application.

These include, amongst others:

• AI-supported cyber offensives,
• misuse of AI for the development of dangerous technologies,
• risks arising from uncontrollable system behaviour,
• automated manipulation of large population groups,
• impacts on critical infrastructure.

Conversely, the mandate will not cover traditional issues of AI ethics, data protection, consumer protection or labour market policy. Other institutions and regulatory bodies already exist for these areas.

Early warning system for policymakers and security authorities

A key task of the institute would be to produce regular situation reports for the Federal Government, security authorities and relevant ministries.

The reports are intended to assess technological developments and analyse potential impacts on national security, the state’s capacity to act and resilience.

This is explicitly not about operational threat prevention. Rather, the aim is to establish a strategic early warning system that supports policy-makers in assessing new technological risks.

Particularly in the field of critical infrastructure, such a capability could become significantly more important in the future. Operators of energy, communications, transport or healthcare infrastructure are increasingly dependent on AI-supported systems and are therefore also exposed to potential new threats.

International networking as the key

Bitkom does not view the institute as a standalone national initiative, but as part of an international network.

Several countries already have comparable institutions, including the USA, the UK, France, Japan, Canada, Australia, South Korea and Singapore.

In the association’s view, Germany should actively participate in international evaluation procedures, security benchmarks and standardisation processes. This also includes cooperation with the European AI Office and international standardisation bodies.

The global nature of AI technologies makes international cooperation a prerequisite for effective security strategies.

The British model as a benchmark

Bitkom cites the British AI Security Institute in particular as a reference.

Within a short space of time, the UK succeeded in establishing an institution that is now regarded as the international benchmark for the technical evaluation of advanced AI systems.

According to the authors, this success is based on several factors:

• strong technical leadership,
• leading international researchers,
• rapid decision-making processes,
• access to high-performance computing infrastructure,
• high scientific visibility,
• close collaboration with leading AI companies.

The institutional approach is particularly noteworthy. The British institute operates less like a traditional government agency and more like a technology-oriented innovation centre within the state structures.

Talent more important than structures

A central argument of the position paper is that the success of a German AI Security Institute does not depend primarily on organisational forms or administrative processes.

Rather, the decisive factor is the ability to attract world-leading specialists in AI security, model evaluation and machine learning.

The authors therefore advocate flexible recruitment models, competitive remuneration and an organisational culture that can keep pace with the dynamism of modern technology companies.

Without top-tier international expertise, it would be virtually impossible to evaluate the most powerful AI models on an equal footing.

Investments in strategic resilience

To establish this, Bitkom proposes start-up funding of at least 100 million euros for the first two years. Thereafter, long-term funding of at least 75 million euros per year should be secured.

The money is to be channelled primarily into personnel, research activities and secure technical infrastructure.

The proposed scale is modelled on the British example and highlights the strategic importance that AI security could have for state resilience and technological sovereignty in the future.

Conclusion

With the proposed German AI Security Institute, Bitkom is painting a picture of a new key organisation in security policy. The focus should be on technological expertise, not regulation.

The idea behind this is clear: anyone wishing to understand and assess the implications of powerful AI systems needs their own scientific and technical capabilities at the very highest international level.

Given the growing importance of AI for cybersecurity, critical infrastructure and the state’s ability to act, the question in future may be less whether Germany needs such an institute – and more whether it can afford to do without it.