Codex Security: now in research preview | OpenAI

I think coding agents are going to be an incredible boost for security. Agents can run as long as you give them tokens and can exercise code in so many ways. Even in my limited projects I've been impressed with security considerations that are brought forward. Perhaps that will be the biggest win is typically developers build the thing and then come back with a security review. Agents tend to think about security while they are building as an incremental aspect.