Swiss AI Safety Days

IASEAI Policy Committee · Red teamer for Anthropic

Advises governments, technology companies, and organizations on the strategy, governance, and responsible deployment of agentic AI systems. She also advises AI Safety Connect and contributes to UNESCO initiatives on AI governance. Previously, she led Global Social Impact Products & Partnerships at WhatsApp and held leadership roles at the World Economic Forum.

Dr. Adrian Hutter leads the Gemini safety post-training team at Google DeepMind. His current research centers on instilling robust values in AI models across diverse environments. He has a background in theoretical quantum physics and has been interested in how to ensure the development of AGI is broadly beneficial for over a decade.

AI researcher and entrepreneur with a PhD in Computer Science from ETH Zurich. LatticeFlow AI is an ETH Zurich spin-off and a global leader in AI security and governance. Gartner recognizes it as a leading vendor in AI Trust, Risk, and Security Management (AI TRISM). Petar previously co-founded ChainSecurity, which PwC acquired in 2020.

Earned his PhD at ETH Zurich in machine learning, symbolic AI and programming languages, and was the sole European recipient of the 2017 Facebook Fellowship. Drawing on more than a decade of research at the intersection of AI, software and formal methods, he develops technology that helps organizations assess and control AI risk with measurable technical evidence.

Previously led UNESCO's work on AI and digital innovation for over 12 years, including as representative to the European Commission, G7, OECD, African Union, Economic Commission for Africa, and Sahel. Sasha received the Public Diplomacy Innovation Prize from the White House under President Obama, and undertook a PhD on responsible AI and the social impact of new technologies.

EU AI Act and OECD AI Expert · IASEAI Council Member

CeSIA is part of the official evaluator consortium for the EU AI Act's Code of Practice on harmful manipulation risks. Initiator of the Global Call for AI Red Lines, endorsed by 12 Nobel laureates and presented at the UN. Founded ML4Good, replicated 20+ times worldwide, and created the EU's first university-accredited course on general-purpose AI safety.

10 Downing Street Innovation Fellow

An AI researcher focused on ensuring AI benefits society. His work spans frontier AI evaluation at the UK AI Security Institute, and public-sector innovation across government, combining technical research with practical policy impact. Also an Honorary Lecturer at Imperial College London, where he teaches deep learning.

Co-founder, AI Safety Lab, KU Leuven

Contributed to shaping the general-purpose AI and systemic risk provisions of the EU AI Act. A PhD researcher at KU Leuven, a member of Estonia's government AI advisory board under the Prime Minister, and co-author of The AI Endgame (Wiley, forthcoming). Also leads the EU AI Act Newsletter, which has over 50,000 subscribers.

Head of the Data Science and AI Lab (dlab)

Works at the intersection of AI, natural language processing and computational social science, with a focus on building AI that is safe for humanity. Previously a Visiting Researcher with the AI Frontiers group at Microsoft Research, and a recipient of the ICWSM Adamic-Glance Distinguished Young Researcher Award and an SNSF Starting Grant.

Formerly AI Safety Research Lead, Palisade Research

Works on an automated red-teaming platform for LLMs and AI agents at Check Point AI Security. Previously led work on specification gaming, cybersecurity agents and malicious fine-tuning at Palisade Research, and collaborated with FAR.AI on testing LLM robustness against multi-stage adversarial attack pipelines. Holds a PhD in mathematics.

Organiser, AI Governance Forum

Leads a Swiss think tank dedicated to governance “of” and “via” AI, a research fellow organization under IASEAI. Specialized in risk management, she works on safety-by-design infrastructure and the governance of agentic AI, has reviewed research projects for the EU Commission since 2018, and holds a postdoctoral fellowship from MIT.

Works on SecureBio's AI team, which focuses on understanding and reducing catastrophic biological risks posed by emerging artificial intelligence, leading targeted evaluations designed to assess model safeguards and identify high-risk capabilities. PhD in Virology from Erasmus MC, Rotterdam, where she studied neuroinvasive mechanisms of high-containment viruses.

Lead, AI Safety and Alignment Group, Max Planck Institute for Intelligent Systems

A member of the EU AI Scientific Panel and a chapter lead for the International AI Safety Report 2026, chaired by Yoshua Bengio, and a founding advisor at Exponential Security Labs. Has taken part in red-teaming efforts for OpenAI and Anthropic models, and co-authored benchmarks used by OpenAI, DeepMind, Meta, xAI and Anthropic. PhD in machine learning from EPFL.

Founder, EuroSafeAI

Also a Research Scientist at the Max Planck Institute in Germany, a member of the EU AI Scientific Panel, a CIFAR AI Chair and faculty at the Vector Institute. Her research focuses on Causal Reasoning with LLMs, and AI Safety in Multi-Agent LLMs. Author of over 100 papers, and a recipient of the ELLIS PhD Award and three Rising Star awards.

Leads a nonprofit R&D lab based in London which provides tools and analysis that help researchers measure the safety of advanced AI.

Leads the team's research and consulting work on the regulation and governance of general-purpose and frontier AI under the EU AI Act. Previously led the European AI Governance team at a nonprofit, and before that spent 15 years as an English barrister and Hong Kong solicitor litigating cross-border commercial fraud in London, Grand Cayman, Guernsey and Hong Kong.

Co-founded a nonprofit institute in Nuremberg investigating how human control of advanced AI could fail and where safeguards could prevent that failure. His research connects threat modelling, safety evaluations and evidence synthesis, focusing on networks of AI agents and pathways to loss of control. Also works on deceptive behaviour and mechanistic interpretability.

His research connects threat modeling and targeted safety evaluations. Inside models, he studies how safety behavior is implemented, for example where a model recognizes harm and where it decides to refuse. On the human side, he asks how routine reliance on AI erodes the competence people need to notice, understand and correct what AI systems do.

Leads research on misalignment and deception in AI agents: how autonomous systems end up pursuing goals nobody intended, how to detect it, and how to fix it. Apollo is an AGI safety research organization building coding agent security products and working with OpenAI, Anthropic, Google, Meta and others on frontier safety, focused on scheming, control and evaluations.

Public Speaker · École Polytechnique

Increasingly focused on AI safety. His YouTube channel Chez Anatole has 40k+ subscribers and 2.4M+ views. Has given 90+ talks across Europe, hosted the International Physics Olympiad, and won the French Debating Association championship at the French National Assembly with École Polytechnique.

PhD Student, ETH Zurich

Researches AI safety, with a focus on developing a rigorous science of misalignment. Alongside his doctorate, he has several years of experience in field-building and AI safety career advising.

Works on making open-weight models safer. A physicist by training, with a master's in Lisbon and a PhD in Theoretical and Applied Mechanics from Sapienza University of Rome, he joined EleutherAI's interpretability team to work on automated interpretability and sparse autoencoders, then moved to data attribution: how specific training data shapes what a model does.

Tobias Häberli directs Pivotal Research, which runs AI safety and biodefense research fellowships in London.

A lawyer holding an LL.M. in technology law from the University of Edinburgh. He participates in the European standardisation groups developing harmonised standards for the EU AI Act (JTC 21) and in the ISO AI committee (SC 42), where he focuses on Frontier AI best practices. Also an external lecturer at Esade, teaching the economics of the AI supply chain.

Founder and Head, XplaiNLP group (JGU Mainz & TU Berlin)

Her research advances explainable, robust, and trustworthy NLP that enables informed human-AI collaboration in high-risk domains, from medical decision support to tools that help journalists and fact-checkers trace disinformation and detect foreign information manipulation. She has raised over €6.5 million in third-party funding and is a GI Junior Fellow.

SteadRise is an AI safety field-building organisation connecting experienced professionals with opportunities in the field. She also collaborates on AI safety community-building and communications, including through the Buenos Aires AI Safety Hub (BAISH). Before moving into AI safety, she spent over 16 years in tech, including six years in enterprise sales at Meta.

Technology operator and impact investor with a career spanning software engineering, scaling startups, and executive leadership at publicly listed global companies, including significant M&A and corporate venture activity.

Co-leads the Talos Fellowship, owns fellowship placement matching, and leads government partnerships. After starting in AI safety community building, she joined Training for Good (now Talos Network) in November 2023 as Employee #1. She is driven by fostering a high-trust, evidence-informed community in European AI governance.

Works on AI as an instrument of influence: how adversaries use it to deceive and defraud people, and how rogue agents might steer the people around them. He builds evaluations of these risks at Apart Research, a non-profit evaluator supporting the EU AI Office. Previously developed misuse and capability evaluations at the UK AI Security Institute and Prolific AI Research.

Works on EU frontier AI governance, focusing on model evaluations, European compute strategy and verification technology. Previously worked on TFS's input to the EU AI Act's General-Purpose AI Code of Practice.

Researches emerging threats to AI agents and develops defenses for agentic systems. With a decade of experience in AI, he began his career building early enterprise AI systems at Swisscom. He later led the machine learning team at RepRisk, before researching vision-language models at Cognex.

Evaluates frontier and open-weight AI models at SaferAI, including its public evaluation of GLM 5.2 against the EU AI Act's Code of Practice. He works directly with Chinese AI developers on risk assessment and safety practice. He also leads a research team studying how training changes the way models represent traits such as honesty.

Works on AI safety and security at Anthropic. Previously, as a fellow at MATS and Anthropic, he worked on AI Control, red-teaming and monitoring. He studied Computer Science at ETH Zürich.

Director of ZAIS, the organization building a central AI safety hub in Switzerland and behind the Swiss AI Safety Days. As an operator, founder and startup advisor in impact and AI, he has built and supported innovation ecosystems across Africa, Asia and Europe. Today, his focus is on helping Europe's strongest technical talent contribute to making AI safe.