I mean it's almost inherent to LLM's right? Like they only know about version before it's knowledge cutoff. I guess it's a big argument for not putting exact versions in files generated by LLM, only major (+minor?)
Yes. My point is that because of training cutoffs it should be mandatory to run SCA scans when dealing with AI code generation.
Not putting exact versions would be a good idea. But that’s not what’s happening today.
Comments
I mean it's almost inherent to LLM's right? Like they only know about version before it's knowledge cutoff. I guess it's a big argument for not putting exact versions in files generated by LLM, only major (+minor?)
Yes. My point is that because of training cutoffs it should be mandatory to run SCA scans when dealing with AI code generation. Not putting exact versions would be a good idea. But that’s not what’s happening today.