I’m personally waiting for legal cases to do with the use of AI trained on code, and whether the licenses apply to it.

If they don’t, our GPL becomes almost useless because it can be laundered but at the same time we can begin using AI trained on code we don’t necessarily abide by the license terms of (maybe even decomps I don’t know how it’ll go). Fight fire with fire and all. So I’d maybe look into that.

If they do, then I’ll probably still use it, but mainly with permissively licensed code, and code also under the GPL (as I use the GPL)

And in both cases, they’d be local models, not “cLoUd” models run by the likes of M$

Until then, I’m not touching it.