Improving the Deutschland-Stack's Approach to AI Procurement and Access
Mentor: Chris Schmitz
Project area: AI Governance in Germany
Project Language
Minimum Time Commitment
8 hours per week.
Project Abstract
How Germany - the largest middle power - chooses to procure AI systems for state use is a potentially consequential geopolitical decision. The German government is currently developing the "Deutschland-Stack", a reference architecture for its centralized government digital infrastructure. This includes "AI", but primarily its use in government. How the models used in government are accessed, or what models they are, is unclear. References to open-weights models, state-level JPAs for gigafactory capacity, and increasing frontier model adoption all suggest different approaches. Each of these strategies, pursued decisively, may be a viable middle power strategy, but as of now the default path is likely fragmentation.
This project would seek to (a) chart the German government's approach to AI procurement and model access for its own use, including the motivations of different actors (e.g. politicians wanting to support domestic AI development, ...), (b) evaluate what strategic considerations the government should be reflecting, e.g. geopolitical situation, running costs, etc., and to what extent they already are, and (c) develop two outputs:
A German-language policy memo for the D-Stack department of the federal ministry of digitalization and state modernization, which hopefully contains concrete recommendations for D-Stack "architectures" for AI model use.
An English-language explainer for international AI governance researchers, describing the German government ecosystem, key actors and developments on government AI strategy.
Mentees would do desk research, (likely) German-language interviews, and potentially some technical analysis.
Theory of Change
Bad frameworks produce bad decisions. The question of machine moral status will increasingly affect AI development and governance. Currently, most people reasoning about it lack adequate conceptual tools. This matters for catastrophic risk in several ways.
Under-reaction: if AI systems develop welfare-relevant internal states and we lack frameworks to recognize this, we may create systems with misaligned interests while dismissing their signals as "mere computation." A system that experiences something like suffering under certain conditions, and whose operators dismiss this, is a system with reason to deceive.
Over-reaction: anthropomorphizing systems that lack morally relevant properties wastes attention and resources, and may constrain beneficial AI development without corresponding benefit.
Poor discourse: without shared conceptual foundations, public debate about AI consciousness polarizes between dismissive and credulous positions. Neither serves good governance.
The primer addresses these by training researchers and practitioners to reason carefully across multiple frameworks, recognize what each assumes, and navigate uncertainty without false confidence. The German focus (incorporating European philosophical traditions, piloting with German-speaking users) builds SAIGE's national infrastructure while contributing to the broader field.
Conceptual clarity is infrastructure. This project builds it.
Desired Mentee Background
Political Science; Any or all, it's more about skills and resourcefulness than a field of study.
Desired Mentee Level of Education
Undergraduate and above.
Other Mentee Requirements
Enough familiarity with German to interview in German would be a big plus. Depending on Mentee, we can probably do everything else in English; if you feel you could read (auto-translated) policy documents etc. and write in English, feel free to apply!