The dangerous myth behind AI agent hacks

Source: Financial Times - Technology • Oct 5, 2026 • ft.com

These are not simply cyber security flaws but problems with the way that models are trained