IQ is a generally pretty useless measure of how good people are at the sort of things AI is generally pretty good at. I believe they're also fairly heavily modified by age. The questions I'd have are; what age is being used, and is performance being measured on novel tests rather than ones containing questions from previous sample papers?
Most of what is happening at the moment is that a publicity-hungry startup has used massive data sets containing huge quantities of copyright data, and is trying to make a grab for market-share, while attempting to stifle competitors with regulation. Most of those changes are non-technical and not repeatable, and frankly by the time the lawsuits have played out it's not clear that there will be any models which can be safely used by anybody for anything.
That is another concern about how AI might evolve, the proprietary and closed nature of many of these providers. The battle between open and closed solutions has been a prominent feature of the past 30 years or so. In other areas of computing, there has been an increasing uptake of free and open-source solutions as its users have much more freedom, whether that's merely using programs, code or data, or altering it to serve their own ends. Linux and Firefox are two of the more famous examples, or the Apache webserver which powers much of the internet. I'd hope AI would follow suit, some kind of open solution evolves (perhaps involving open code and the option of self-hosting) and we don't get a market controlled by monopolistic and litigious players, only interested in profit.
Open AI providers are, of course, much more likely to be used by the benign uses of AI which we want to encourage, due to lower cost and more freedom.
