That's pretty much why gatekeeping security, as Anthropic and OpenAI want, isn't a thing and doesn't work.

All projects need security and writing code means taking security decisions, you can't untangle both.