Simon Jarvers

PhD researcher in AI governance at TU Munich.

About

AI is a transformative, general-purpose technology that we need to make safe and beneficial for everyone. I am currently interested in two areas: how to govern AI well, and understanding its impact on society.

I am a PhD researcher at the Professorship for Societal Computing at TU Munich. I also work embedded as an AI governance specialist at an AI startup, which lets me study governance from the inside through action research.

Outside of research, I love spending time in nature, building my own furniture, and trying to beat my personal records as a hobby powerlifter.

Research

Scalable Governance

I am building Scalable Governance, a research initiative with the goal of scaling AI governance to the pace of rapidly increasing capability and volume of AI systems. My research so far studies this at the level of the organisations that build AI: how teams turn regulation into practice, who shapes what compliance means, and how AI agents can do governance work under human oversight. More information will follow once the initiative is founded and public.

  1. Human-AI Collaboration in Compliance AutomationDesigning an Agentic Knowledge Base for AI Governance Forthcoming HICSS 2027
    • Simon Jarvers
    • Carolin Röthemeier
    • Holger Wittges
    • Orestis Papakyriakopoulos

    Proceedings of the 60th Hawaii International Conference on System Sciences (HICSS)

    AI governance work draws on a living body of organizational knowledge, policies, and system facts that drift and go stale as organizational and regulatory contexts change. For resource-constrained organizations under the EU AI Act and related standards, maintaining this knowledge consumes scarce compliance capacity. We applied a pattern drawn from practice, in which an LLM agent maintains a structured, citation-anchored knowledge base for AI governance. Through a Design Science Research study at an AI startup, we apply the artifact to two ISO management system standards. The artifact separates a human-authoritative source layer from an agent-moderated knowledge layer in the company’s internal governance work processes. We abstract two design principles, human authority over agent moderation and traceability, which demonstrate how to responsibly leverage advancing AI capabilities for human-agent collaboration in regulated compliance settings.

  2. Governing by ProxyHow Regulatory Intermediaries Shape EU AI Act Compliance in Practice AIES 2026
    • Simon Jarvers
    • Orestis Papakyriakopoulos

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society 9(2), 1373–1384

    The EU AI Act assigns compliance obligations to defined actor roles along the AI supply chain, but how these obligations are enacted in practice remains understudied. This paper investigates the ecosystem of regulatory intermediaries (consultants, auditors, ethics advisors, certification bodies) through which an AI system provider translates regulatory requirements into organisational practice. Drawing on participant observation and document analysis from an embedded case study at an AI startup, the analysis applies the Regulator-Intermediary-Target (RIT) model to map sixteen intermediaries and trace the compliance pressures that drive pre-enforcement intermediary engagement. Compliance artefacts across risk management, impact assessment, and technical documentation show a consistent pattern: intermediary interpretive choices construct what AI Act compliance means for the regulated organisation. Using Edelman’s legal endogeneity framework, which theorizes how organisations’ private compliance constructions can feed back into the meaning of law itself, we discuss how legal requirements are managerialised into procedural deliverables and how compliance forms diffuse without evidence of governance efficacy. The paper contributes on two levels: it extends the RIT model to organisational-level AI governance, demonstrating that compliance intermediaries perform constitutive rather than facilitative functions, and it provides an early empirical signal for legal endogeneity in AI Act compliance. We encourage more guidance from the EU and argue that the compliance ecosystem deserves the same analytical and regulatory attention currently directed at the AI supply chain.

  3. Governance-Driven DevelopmentEmbedding Regulatory Requirements in AI Development Workflows ECAF 2026
    • Simon Jarvers
    • Orestis Papakyriakopoulos

    Proceedings of the Fifth European Conference on Algorithmic Fairness (ECAF), PMLR 350, 365–371

    AI governance under the EU AI Act faces a structural misallocation of effort: management-based regulation, combined with the credence-good properties of governance outcomes, pushes resource-constrained organisations toward verification-facing documentation rather than the implementation it is meant to verify. We propose Governance-Driven Development (GDD), a framework that treats AI Act obligations and the requirements of supporting standards for safe AI development as formal specifications for software development work. GDD runs on two nested action cycles. A planning cycle scopes obligations against organisational context to produce specifications. An implementation cycle converts each specification into the artefact it specifies, with traceability back to the regulatory source. AI agents perform template-driven work; humans retain interpretation and accountability. The framework is in early implementation at an AI Startup, with unresolved load-bearing questions around knowledge base architecture and obligation decomposition.

  4. Engaged AI GovernanceAddressing the Last Mile Challenge Through Internal Expert Collaboration FAccT 2026
    • Simon Jarvers
    • Orestis Papakyriakopoulos

    Proceedings of the 2026 ACM Conference on Fairness, Accountability, and Transparency (FAccT), 4415–4438

    Under the EU AI Act, translating AI governance requirements into software development practice remains challenging. While AI governance frameworks exist at industry and organizational levels, empirical evidence of team-level implementation is scarce. We address this “Last Mile” Challenge through insider action research embedded within an AI startup. We present a legal-text-to-action pipeline that translates EU AI Act requirements into actionable strategies through internal expert collaboration by extracting requirements from legal text, engaging practitioners in assessment and ideation, and prioritizing implementation through collective evaluation. Our analysis reveals three patterns in how practitioners perceive regulatory requirements: convergence (compliance aligns with development priorities), existing practice (current work already satisfies requirements), and disconnection (requirements perceived as administrative overhead). Based on these patterns, we discuss when governance might be treated genuinely or performatively. Practitioners prioritize requirements that serve end-users or their own development needs, but view verification-oriented requirements as box-ticking exercises. This distinction suggests a translation challenge: regulatory requirements risk superficial treatment unless practitioners understand how compliance serves system quality and user protection. Expert collaboration offers a practical mechanism for transforming governance from external imposition to shared ownership and making previously invisible governance work visible and collective.

  5. Participatory AI and the EU AI Act AIES 2025
    • Chiara Ullstein
    • Simon Jarvers
    • Michel Hohendanner
    • Orestis Papakyriakopoulos
    • Jens Grossklags

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society 8(3), 2550–2562

    Participatory AI calls for the involvement of stakeholders in AI design, development, evaluation, and deployment to attain more inclusive, transparent, and accountable AI. However, actual implementations of participatory AI remain little incentivized by governments, despite appeals issued by academia and also industry. In this work, we investigate the role of ‘participation’ in the obligations of AI system providers and deployers set out by the EU AI Act. First, we analyze the gaps between the participation explicitly stated in the non-binding recitals of the AI Act and the provisions of the Act itself, showing that the legal demand for participation is limited. For example, neither Article 9 on risk management systems nor Article 27 on the fundamental rights impact assessment mention any form of participation. Article 95 on the voluntary codes of conduct is the only enacting term that explicitly suggests stakeholder participation. Second, based on these results, we analyze opportunities for participation emerging from the obligations of high-risk AI system providers and deployers (AI Act, Chapter III, Sections 2 and 3). We identify five clusters of obligations with participatory opportunities: risk management, data and data governance, information provision, resilience testing, and impact assessment. Third, we provide examples of use cases for each of the identified opportunities for participation. This work contributes to a better understanding of regulatory demands and practical opportunities regarding participatory AI in the context of the AI Act.

  6. Uncertainty as a Primary Barrier for Trustworthy AI Under the EU AI ActGerman SME Perspectives EWAF 2025
    • Simon Jarvers*
    • Chiara Ullstein*
    • Jens Grossklags

    * Both authors contributed equally.

    Proceedings of the Fourth European Workshop on Algorithmic Fairness (EWAF), PMLR 294, 281–287

    The European Union’s AI regulation, the EU AI Act, represents a significant shift from voluntary ethical frameworks to binding regulation, presenting implementation challenges particularly for resource-limited SMEs. Our mixed-methods research examined the EU AI Act’s impact on SMEs through surveys of German AI SMEs (N = 21) and interviews with AI SMEs and industry stakeholders (N = 13). In this extended abstract, we summarize our motivation and methods, and focus on providing results from the interviews. Our findings reveal that company size and compliance experience significantly affect estimated implementation capabilities. SMEs face considerable resource constraints across time, finances, and staffing. Implementation uncertainties – including definitional ambiguity, unclear scope, and insufficient guidance – drive strategic responses: delaying compliance efforts, modifying products to reduce regulatory burden, and frequently seeking external compliance expertise and certification. These results indicate that uncertainty emerges as the primary implementation barrier. Researchers can help reduce uncertainty by developing best-practice guidelines that support the AI Act’s trustworthy AI objectives. We conclude with recommendations for policymakers and researchers.

AI in research

Exploratory

How AI changes knowledge work and its impact on epistemology in academia. A first study, on how researchers disclose their use of generative AI in published papers, is currently under review.