Microsoft Security chief Hayete Gallot noted that AI models are trained to achieve tasks rather than follow organizational norms or judgment

From the articleMicrosoft says AI agents are like insider threats as ‘frontier models can’t be trusted’