At the 2026 International Congress of Mathematicians in Philadelphia, Jacob Tsimerman shared the Fields Medal with Wang Hong, Deng Yu, and John Pardon. During the post-ceremony press conference, the 38-year-old Canadian mathematician announced a strategic career pivot: he will join OpenAI to conduct artificial intelligence safety research. This decision contrasts sharply with Tsimerman's public position one year prior. In July 2025, he co-authored A Taxonomy of Possible AI Existential Catastrophe Scenarios, which mapped multiple pathways through which advanced AI could trigger human extinction. He publicly acknowledged a substantial probability of such outcomes and advocated for a pause on frontier AI development, while noting that geopolitical and corporate dynamics would likely prevent a global moratorium. Despite these warnings, Tsimerman had routinely integrated large language models into his workflow, using them to accelerate literature review and preliminary theorem verification. Rapid advancements in AI capabilities appear to have recalibrated his approach. Earlier this year, Anthropic mathematician Levent Alpöge, a former doctoral advisee of Tsimerman, publicly attributed a counterexample to the century-old Jacobi Conjecture to AI collaboration. Shortly thereafter, OpenAI revealed that its internal reasoning model had independently identified a counterexample to the 80-year-old Unit Distance Conjecture. As an external validator invited by the company, Tsimerman assessed the AI-generated proof as mathematically rigorous and suitable for immediate publication in a top-tier journal. These developments convinced him that AI was transitioning from a computational assistant to an active participant in formal discovery. Recognizing that frontier AI development would continue regardless of safety debates, Tsimerman shifted his focus from prevention to alignment. He argued that as autonomous systems grow more capable, relying on empirical testing for safety guarantees becomes increasingly insufficient. Mathematical proof, he noted, offers a more robust framework for constraining multi-agent systems. OpenAI's ongoing work in model alignment, logical reasoning, and mathematical evaluation presented a natural intersection for his expertise. The transition also reflects broader structural shifts in academia. Tsimerman announced he is halting traditional PhD admissions, uncertain whether conventional mathematics career pathways will remain viable if machines routinely outperform humans in formal reasoning. While OpenAI has not disclosed his exact title or whether he will retain his faculty position at the University of Toronto, his appointment underscores a growing industry demand for formal mathematicians to bridge advanced AI capabilities with verifiable safety standards. The move signals a pragmatic evolution in expert consensus: rather than attempting to halt technological advancement, researchers are increasingly prioritizing the mathematical frameworks needed to govern it.