Smart people.
Maybe average users will also limit their ChatGPT use -- not because of "data leaks", but because...I dunno, because the idea of getting your knowledge from an unfeeling, emotionless AI that simply regurgitates random information from the internet with supreme confidence and encourages you to accept that wisdom as gospel without questioning it or thinking critically simply reinforces the absolute worst trends in society that have already been exacerbated enough with people just streaming endless bullshit from social media into their brain is scary as absolute **** and quite possibly, non-hyperbolically represents the downfall of humanity.
...But yeah, probably not.
I feel like the way people use tech has always been a steady descent towards our own downfall in one way or another, going all the way back to advancements in communication in one-to-many media formats like newspapers and radio. Now we've moved to a mixture of one-to-many and many-to-many, as individuals can now direct feedback towards media outlets and talk with one another regardless of physical distance. Certain voices rise above the noise to become leading voices or banner-wavers of one kind or another, and what we're seeing is that those who do rise to prominence aren't necessarily those with the most accurate information, most correct analysis, or most level-headed takes... but the ones that drive emotional responses, because the new currency of the internet is attention. There are things that garner positive attention, but they usually fall way short of the content that drives anger and outrage. To play both sides of this engagement algorithm, content creators have been focusing on polarizing topics and using divisive speech to attract positive attention from one side, and anger and derision of the other. Because the most important thing now is that people engage your content, and the algorithm doesn't care whether it's being shared for positive reasons, or out of outrage.
The way social media has been weaponized since before the 2016 US election, particularly the use of SNS data by Cambridge Analytica to influence public opinion, signaled to me that data was no longer a reflective tool for understanding trends and public sentiment, but that the street went both ways, and the same tech could be used to affect real-world outcomes.
AI is very much just another grease stain down this slippery slope, perhaps even representing a ramp of some sort, giving humanity a chance to catch some sick air before we come crashing down. Bots and troll farms have already been deployed to create the illusion of public sentiment, and these new forms of AI are likely to make it easier to synthesize and randomize this kind of spam content, pretending to be actual users, to increase the noise in the system and squeeze out actual, rational human voices.
The problem as I see it doesn't come from the existence of the tool itself, but the way it is used, and also the fact that we as humans are prone to anthropomorphize things and project humanity onto them. For some reason, people want to engage with sentient life so much that we'll talk to AI (in its current iteration) as if we're talking to a person, rather than a tool. To me, ChatGPT and its fellows are just slightly more advanced search engines, capable of synthesizing the content we're searching for rather than pointing us to the sources where it can be found. I think the real danger lies in the black-box nature of the technology: whether it's because the companies behind AI don't realize the importance of transparency, or they're protecting proprietary technology, I think it's super valid to be suspicious of the information AI churns out because you don't know where it's coming from; there's no way to check the sources and even see if the response is valid, AI will pump this stuff out with an alarming level of confidence, and not only does it take a discerning eye (and someone who knows enough to know what's incorrect) to spot the errors. At this point, I can still look at a block of text and generally tell whether it's AI generated (we are pattern-recognizing creatures, after all), but I think a lot of people aren't scrutinizing the information they're absorbing nearly as much... and this creates new opportunities to fool and control people in nefarious ways. I feel that the future is fraught with all kinds of novel dangers we're creating for ourselves.