Back to News
Release2026-03-11
Designing AI agents to resist prompt injection
Source: OpenAI
How ChatGPT defends against prompt injection and social engineering by constraining risky actions and protecting sensitive data in agent workflows.
openaigptagentsprompting