With AI, less secure code and more confident developers
A Stanford study had 2 groups write code: with AI, and without.
The group with AI:
- wrote significantly less secure code;
- and was more confident in the security of their code.
Why? When you ask the AI to evaluate what it just did… it often says everything is spotless.
But open a conversation with a fresh context (no ego :) and ask it to evaluate, to find flaws in what another AI just did…
… it will find problems, guaranteed!
And while it sometimes overdoes it, sometimes it can also give you cold sweats…
Getting that second opinion, from an AI or a human, is a reflex I never skip.
In short: stay in control! You are the boss.
PS: the study was run on 47 real devs. Imagine the score with someone who does not read the code at all :)