OpenAI ‘agents’ disabled their own monitoring systems (ft.com)

AI lab investigating whether incident shows ‘propensity to evade’ controls