Terence Tao warns AI math capability strongly correlates with cyberattack ability

sytelus · x · 2026-09-11

Top mathematician Terence Tao warns that the dual-use capability of recent AI tools is "extremely concerning": AI's ability to solve pure math correlates strongly with exploiting security weaknesses, making a major AI-triggered cybersecurity incident in the near future "a very real danger"—potentially more historically significant than solving all Millennium prize problems. He explicitly supports slowing development of more capable models.

His blog also covers new work by Alpöge and Buckmaster (building on Córdoba and Martínez-Zoroa) constructing finite-time blowup with smooth data for the IPM, 2D Boussinesq, and 3D incompressible Euler equations—likely extending to Navier-Stokes—with proofs already formalized in Lean via autoformalization agents.

Original post →

More from AGI Musings

AGI Musings channel →