My current research asks what happens to consent, oversight and accountability once AI systems stop answering questions and start acting on people’s behalf, increasingly in interaction with other agents.

Work in this line formalises consent chain degradation in embodied multi-agent systems and proposes a runtime architecture for consent verification (ICRA 2026 Workshop on Robot Ethics); analyses the consent vacuum that emerges in open-ended agent-to-agent ecosystems and maps it onto the EU AI Act, the GDPR and the Digital Services Act (IAIL 2026); examines why consent breaks when AI moves from chat to action (CHI 2026 workshop); proposes accountability graphs for tracing responsibility across conversational AI chains (ACM CUI 2026); and offers a sociotechnical reading of the Clawdbot/Moltbook case (İstanbul Üniversitesi Sosyoloji Dergisi, 2026).

The research connects to my policy work on trustworthy AI, including contributions to the OECD Framework for the Classification of AI Systems and OECD work on accountability in AI, and to earlier papers on AI content detection obligations and fairness in medical AI.