A mathematician who just won the field’s highest honor—and who co-authored a paper on AI-triggered human extinction—is now working at the company most responsible for pushing the technology forward. Jacob Tsimerman, the University of Toronto number theorist who received the Fields Medal this year, has joined OpenAI’s safety team. The move puts a leading voice on existential risk inside the very organization building the systems he’s warned about.
Tsimerman doesn’t mince words about the stakes. AI remains “mostly on an empirical basis with few guarantees about how these systems actually work,” he said. That’s a problem when you’re deploying technology he believes could contribute to “omnicide events.” Panic isn’t useful, he argues, but honest risk assessment is non-negotiable. His presence at OpenAI signals the company is taking mathematical rigor seriously—or at least wants the credibility that comes with hiring someone who does.
His timing is notable. Former DeepMind CEO Demis Hassabis recently weighed in on AI’s mathematical capabilities, calling current advances impressive but short of a true breakthrough. The bar, he says, is something like AlphaGo’s legendary “Move 37”—a genuinely creative leap. Today’s models, including OpenAI’s new Astra system, still stumble on Millennium Prize Problems. Tsimerman is convinced that won’t last. He expects AI to surpass human mathematicians entirely, and soon.
What changes when the researcher most worried about extinction works alongside the engineers racing toward superhuman intelligence? Tsimerman thinks mathematicians can contribute something the field desperately needs: rigorous guarantees about how these systems behave. The question is whether anyone’s listening fast enough. He joins at a moment when safety researchers have been leaving OpenAI in clusters, often citing frustration that product launches outpace caution. His bet appears to be that influence from the inside beats criticism from the outside—a wager the rest of us have no choice but to watch play out.
