Be careful evaluating code from LLM based tools, as there are several avenues for malicious users to inject output:
miniblog.
Related Posts
A new class of typosquatting attacks for malicious packages: register package names that are hallucinated by ChatGPT: https://vulcan.io/blog/ai-hallucinations-package-risk
(h/t @rauschma)
Malicious ads discussion on TV Tropes https://tvtropes.org/pmwiki/posts.php?discussion=13223684920A10189100 discusses the importance of antivirus software.
I suspect that having an up-to-date browser is more important these days? It's probably the most attacked software by far.
How do you prevent "trusting trust" attacks with malicious compilers?
You don't need a trusted production-grade compiler. It's sufficient to have a really limited trusted compiler or even an untrusted compiler provided the triggers don't overlap.
