They're running code on users' computers that it would not be reasonable to think that the user consented to running on their computer. This is CFAA-violation-shaped. Of course, they won't be prosecuted if it is indeed a violation, and I do not know for sure if it meets the specific legal criteria. However, it is something that I think should be illegal. Make it so if software does something that would be unreasonable to think that the user wants to happen, it needs to make that abundantly clear before it does it, otherwise it's a CFAA or similar violation. This would, of course, have very broad consequences. However, this Claude issue feels particularly violating to me.