On 3/23/26 07:33, Theo de Raadt wrote:
>> I understand the code the AI provided could be stolen and you don't
>> want to play that game, ok.
> 1) It is blatantly obvious to everyone that these AI were provided with
>     large volumes of code, and that licenses on the files are ignored,
>     because from the companies's perspective this is "teaching".
>
> 2) The word "stolen" is being used incorrectly.  Nothing is stolen.
>
> 3) The output of these AI is a sophisticated form of pattern regurgitation.
>     Regurtigating the old.  Generally the regurgitation is differnet.
>     But that's not the real problem.
>
> 4) Noone can put a Copyright onto that output, because under the current laws
>     that requires a human to have created the output.

> Lacking Copyright (or similarily a Public Domain declaration by a human), we
> don't receive sufficient rights grants which would permit us to include it
> into the aggregate body of source code, without that aggregate body becoming
> less free than it is now.

The 4th point is an interesting issue not only for copyright in general. 
We face a growing lack of understanding the machinesright also in 
respect of letters of law. We can see that people use words like 
"learning" and "thinking" when using/describing/interact with LLMs/AI 
which is a fundamental misunderstanding of whats going on and make it 
harder to clearly judge the nature of the output of any LLM.

So one question (we need to get an answer for) might be when is the 
point we would treat any given output as enough created/influenced by a 
human. Using code snippets from googleAI or copilot seems to be ok for 
now (for most peoples). How we treat wipe coded software - when is it 
"viewed (perceived) sufficiently" to accept it as human work or even 
treat it equally?

best, Matthias