TSG236

@GrapheneOS
1. To what extent did you use LLMs in both the coding and review process? What are the lines you are not willing to cross?
2. How much did you pay for the models, and to which company/provider did you pay it to?
3. Are there any considerations on hiring more dedicated vulnerability researchers instead of spending on LLMs?
4. How do you justify the harms stated above?
5. Finally, how do you plan to mitigate the harms of using LLMs? Are there plans to use it locally, for instance?

@GrapheneOS I'm no purist here, I've used LLMs. (Just barely though, I've hardly found it useful for the things I do.)

I see its potential utility, and I do wish to see that this technology gets better, both in capability and in ethics. So, in the interest of transparency, I have a few questions:

@GrapheneOS You guys have made it very clear that you find frontier LLMs useful for this specific purpose, and that LLM contributions that are slop are not accepted. I trust you to know what is best for the future of the project I've followed since the Copperhead days.

However, my main concern regarding this is the legitimization of a very environmentally damaging service (not to mention the direct support of fascists) by endorsing its use.