Arent't the models trained on open source code though? In which case OpenAI et al should be following the licenses of the code on which they are trained.
From a common FOSS contributor license...
>>permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions...
https://opensource.org/license/mit
... As opposed to a visual artist who has signed away zero rights prior to thier work being scraped for AI training. FOSS contributors can quibble about conditions but they have agreed to bulk sharing whereas visual artists have not.
Stealing from FOSS is awful, because it completely violates the social contract under which that code was shared.
I don't know what you mean by a rugpull exactly, but of course in theory you can grant/obtain very extensive rights under a CLA as well, including eg the permission to relicense your contributions under whatever terms the licensee prefers. CLAs are a great way to centralize the IPR in an open source project for practical purposes like license enforcement, but in case the CLA terms allow it, the central governing entity could also obtain the right to switch the license even to a, say, commercial one. (Such terms would usually be a red flag for contributors though.) And in any case, that kind of CLA wouldn't still close off the code already released under the previous open-source license, and neither would it prevent you from licensing your own contributions under terms of your choice.