Well, whatever use you're putting this tool to, your 'adversaries' or those who want to thwart your efforts will be using the same tools. If they have access to more powerful AI tools than you do and are willing to use them for malevolent means, guess who wins. That's an oversimplification, obviously, but arguably it's much easier to use a very powerful tool for attack and destruction than for positive purposes.

For one thing, AI will make surveillance much easier and will take it to a level not possible without it. For example, let's say someone has 10 different accounts on Reddit or whatever platform and uses them all for different purposes in the hopes of remaining anonymous. Let's say they use a VPN to further mask their identity. Pre-AI, it was very difficult to tie these accounts together, but with AI, it will be simple. Your language style is like a fingerprint. If the ORG GODS already have good samples of your language use, my guess is it will now be easy for them to identify your 'anonymous' accounts. In short, anonymity is now impossible.

It won't just use language. It will use time of posting, VPN IPs, interlocutors, themes of messages, place names mentioned, age (easy to nail, I suspect), gender—basically, all available data can now be crunched in seconds rather than requiring a room with eight people and bad lighting, hunched over computers and getting carpal tunnel syndrome as they try to connect the dots.

But that's just surveillance. The big threat here is mind control via media. The Internet has become a collection of a very few portals run by the ORG GODS, and it is now a weapon more than anything good. Pre-AI we were fucked. Now we're super fucked, like so fucked we can't unfuck ourselves, and any effort to unfuck ourselves will only make us more fucked. It's like we're in a Mitchell and Webb skit.

To avoid surveillance if you're trying to stay anonymous online, one thing one could try would be to take their messages and run them through an anonymous, encrypted AI engine like Proton's Lumo, telling Lumo to take your message and reword it in some way, like it was written by a native Spanish speaker who learned English late in life, or someone from a different socio-economic-geographic background, etc. It's very good at doing that.

If a student wants to avoid detection of their use of AI, here's what I would recommend. Let's say they need to write an essay on the first moon landing. OK. Tell AI to write the essay. 500 words, bla bla bla. Now give AI a sample of your own, original writing. The more samples, the better. Past essays that they actually wrote would be great. Now tell AI to take that essay it just wrote and redo it in the style of your own writing, grammar mistakes and all. Bingo. Take that teacher. I'm so glad I retired.