Can confirm they patched it and I tried a few different methods. That being said, with the way all this stuff works I can see 2 cases:
They just hard coded it to shut down whenever the user prompts it with some combination of "Windows" and "Keys", in which case Chat GPT can still be exploited in similar ways for a ton of other fun piracy uses.
They made it "intelligently" detect when the user was trying to trick it, which is what it (nominally) has always tried to do, so there's still a billion ways to get it to give away sensitive info because AI don't real.