Proof Against Attack: Can AI Help Build Software We Can Trust?

AI is making cyberattacks easier to launch, harder to spot, and possible at much greater scale. At the same time, more of everyday life depends on software: hospitals, energy grids, transport systems, financial services, and the devices we carry in our pockets. This talk asks a simple question: what would it take to make the software we rely on dramatically safer? One answer comes from formal methods: techniques that use mathematical proof to check that software behaves as intended. Until now, these methods have often been seen as too specialist, too expensive, or too slow for widespread use. But AI may change that. The same advances that could help attackers, could also help us find weaknesses and remove them before systems are deployed –  creating a unique opportunity to build toward a world without software vulnerabilities.Speaker: Kathleen Fisher, Advanced Research and Invention Agency, UK Part of CAV and FLoC 2026.