Systemic Risks from General-Purpose AI

Описание: Abstract: This talk will first give an overview of systemic risks from general-purpose AI under the EU AI Act, then present early findings from research on a taxonomy of systemic risks, and finish with ideas on how to mitigate such risks. Using a systematic review of 1,781 scholarly works, we selected 86 papers that yielded 13 risk categories and 50 contributing sources. Systemic risks – defined by the EU AI Act as large-scale threats to societies or economies – span environmental damage, structural discrimination, governance failures, and loss of control. Prominent drivers include knowledge gaps, difficulty recognizing harm, and the unpredictable evolution of AI systems. To assess mitigation measures, we surveyed 76 experts from AI safety, critical infrastructure, democratic governance, CBRN, and bias disciplines. From 27 literature-derived mitigation measures, three emerged as both highly effective (expert agreement above 60%) and technically feasible: (1) safety incident reporting and security information sharing, (2) third-party pre-deployment model audits, and (3) pre-deployment risk assessments. Experts emphasized external scrutiny, proactive evaluation, and transparency as core principles.

Speaker Bio: Risto Uuk is the Head of EU Policy and Research at the Future of Life Institute in Brussels. He contributed to the collaborative, evidence-driven process that led to the inclusion of general-purpose AI and systemic-risk provisions in the EU AI Act – the world’s first comprehensive AI-governance framework, designed to foster responsible innovation in Europe. He’s currently co-authoring a book, The AI Safety Endgame, with Professor Lode Lauwaert (Wiley, forthcoming) – a strategic guide to long-term AI safety. As a PhD Researcher at KU Leuven, Risto focuses on systemic risk assessment and mitigation for general-purpose AI systems. He recently served as a Visiting Researcher at Stanford’s Digital Economy Lab, examining the economic and societal implications of advanced AI. He founded and leads the biweekly EU AI Act Newsletter with nearly 50,000 subscribers, and created a website that ranks among the top search results for the AI Act globally. Previously, Risto worked for the World Economic Forum on a project about positive AI economic futures together with Stuart Russell, Daniel Susskind and others, and did research for the European Commission on trustworthy AI. He holds a master’s degree in Philosophy and Public Policy from the London School of Economics and a bachelor’s degree from Tallinn University. Risto was selected as a Fellow of the International Strategy Forum, an initiative by former Google CEO Eric Schmidt (in partnership with European Council on Foreign Relations), and received the Global Priorities Fellowship from the Forethought Foundation for Global Priorities Research.