news
Designing AI agents to resist prompt injection
March 11, 2026
ChatGPT defends against prompt injection and social engineering by constraining risky actions and protecting sensitive data in agent workflows. The key point is that agentic systems need guardrails around tool use and data access because malicious prompts can otherwise steer models into unsafe behavior or leaks.
How ChatGPT defends against prompt injection and social engineering by constraining risky actions and protecting sensitive data in agent workflows.
Source: openai.com