Current Position & Biography Yoshua Bengio is Co-President and Scientific Director of LawZero, a Montréal-based nonprofit researching safer advanced AI. His current institutional biography also identifies him as a full professor at Université de Montréal, founder and scientific advisor of Mila, and chair of the International AI Safety Report. These roles place him at the intersection of machine-learning research, research institution building and public discussion of AI risk.
LawZero was founded in 2025 following a change in his research priorities toward AI safety. His current Mila title is scientific advisor, not the scientific-director title found in older biographies. Coverage should distinguish his own policy and scientific arguments from conclusions endorsed by every institution or report contributor with whom he works.
Notable Contributions Bengio shared the 2018 ACM A.M. Turing Award with Geoffrey Hinton and Yann LeCun. ACM recognised their conceptual and engineering contributions to the development of deep neural networks. That award documents foundational research influence; it is not certification of a later product, policy position or proposed safety architecture.
LawZero’s research centres on Scientist AI, an approach intended to separate prediction and understanding from a system’s pursuit of its own goals. Its July 2026 featured paper, Safety from Honesty in a Disinterested AI Predictor, presents a formal safety argument under stated assumptions about training and the prevalence of dangerous predictors. The distinction between a mathematical argument and a deployed system is important: the research does not establish that arbitrary AI systems are safe, and the authors acknowledge that a predictor can be incorporated into an agentic system.
In September 2026, LawZero reported a joint funding commitment of up to CAD300 million from Canada and Germany. It described plans for research growth, computing infrastructure and a Berlin office. This is an announced upper-bound commitment, not a statement that the full sum has been disbursed or that the planned infrastructure is complete. Funding provides capacity to test a research programme; it does not validate its conclusions.
Why Katharos Is Following Them Bengio offers a perspective on AI safety grounded in a career that helped develop the underlying technology. His work raises a concrete question: can useful capability be expanded while restricting the incentives and opportunities for a system to act autonomously? The answer requires evidence about actual training procedures and deployment conditions, not just compelling terminology.
Areas to Follow Scientist AI research; independent evaluation of safety arguments; evidence about agentic behaviour; public research funding; and the relationship between AI capability and controllability. Particular attention should go to which assumptions survive implementation and how researchers report limitations or unsuccessful tests.
Reporting Scope Follow attributable research, institutional developments and public arguments. Distinguish anticipated catastrophic risks from observed outcomes, theoretical guarantees from empirical results and advocacy from consensus. This profile neither predicts a particular AI catastrophe nor endorses claims that a single architecture has solved AI safety.
Katharos Category Artificial Intelligence
Photo Credit Archival portrait: “Yoshua Bengio 2019”, Maryse Boyce, 18 March 2019. CC BY 4.0; cropped and resized. Source: https://commons.wikimedia.org/wiki/File:YoshuaBengio2019.jpg ; license: https://creativecommons.org/licenses/by/4.0/
