**OpenAI: Prompt Injection Defense**

This is a blog post from OpenAI (March 11, 2026) discussing how to design AI agents to resist prompt injection attacks. It argues that as AI agents become more capable, prompt injection attacks increasingly resemble social engineering rather than simple instruction overrides. The post suggests that defense should focus on constraining the impact of manipulation rather than solely filtering inputs, and highlights that traditional 'AI firewalling' approaches are insufficient against sophisticated attacks that use social engineering tactics.