Hacker News new | ask | show | jobs
Aardvark: OpenAI's agent security researcher (openai.com)
1 points by taocp 234 days ago
1 comments

It lacks a systematic theoretical framework to support it, as there is no guarantee regarding the detection rate of vulnerabilities. However, its practical effectiveness is undeniable—it can definitely identify vulnerabilities and achieve a relatively low false positive rate. In a nutshell, it is a "nice-to-have" tool. It would be valuable if someone could conduct a capability and ROI comparison between this tool and traditional security detection tools.