She led the establishment of CARMA’s Comprehensive Risk Assessment programme and the research behind the framework introduced in the paper Adapting Probabilistic Risk Assessment for AI. This work has informed the EU AI Act Code of Practice. Her broader research treats advanced AI as a sociotechnical system, applying systems thinking to risks ranging from systemic failures and loss of control to effects on institutions, epistemic infrastructure, and human autonomy.
Anna is a member of the EU’s Scientific Panel of Independent Experts in AI, established under the AI Act to advise the AI Office and national authorities on general-purpose AI and systemic risk.
Previously, she was a Visiting Scholar at the Leverhulme Centre for the Future of Intelligence at the University of Cambridge and an AI Consultant at McKinsey QuantumBlack, and was acknowledged in OpenAI’s GPT-4 system card for research informing its risk assessments. Earlier, as a machine-learning engineer at Ericsson, she worked on user mobility prediction, new fault-type identification, and drone airspace management. Her research in quantum error correction included an error-rate-agnostic decoder for topological codes, published in Physical Review A. She holds a BSc in Engineering Physics from Chalmers University of Technology, where she also pursued MSc studies in Complex Adaptive Systems.