updateMar 11, 2026ยท 1 min read

Designing AI agents to resist prompt injection

How ChatGPT defends against prompt injection and social engineering by constraining risky actions and protecting sensitive data in agent workflows.

chatgpt has implemented new defenses against prompt injection and social engineering. these measures constrain risky actions and safeguard sensitive data within agent workflows.

for game developers using chatgpt, this means improved reliability and security when integrating the model into projects. the updates help ensure that agents behave as intended without being manipulated by harmful prompts.

there are no specific pricing changes associated with this update. developers can continue to use chatgpt with enhanced protections in place.

vibe check
chatgpt's new defenses mean your agent slightly harder to jailbreak, which is good because you definitely weren't going to implement auth properly anyway