Well now, if AI thinks someone is dead when they're demonstrably not...
I'm not confident that this is fit for purpose.
See my earlier posts:-
https://forums.theregister.com/forum/all/2023/07/26/openai_ai_classifier/
The US Internal Revenue Service has said it will use AI software to go after wealthy individuals and corporations violating tax laws. The tax agency said experts in data science and tax enforcement have been applying "cutting-edge machine learning technology" to find anomalies in partnership taxes, general income taxes and …
Although Death and Taxes are frequently mentioned together in practice what tax organizations like the IRS regard as "death" isn't quite the same thing as the biological phenomenon. Put another way, just because you're dead doesn't mean your existence ceases -- the tax people may well have a claim on your estate and its likely to be well up the priority list (and unlike normal creditors you have to prove you don't owe them any money rather than they have to substantiate their claim).
The rest is obvious. They're going to go after the better heeled because a) its because where the money is and b) they're the ones that employ various 'tax saving strategies' because they've not only got the assets to protect but also the wherewithal to protect it with.
...bit of case law out of Delaware. Basically is says that in the US publicly traded corporations only exist to generate profit for the holders of said corporations' stock. So if the CEO want to try to blow smoke up my ass about how much Google wants to make my life better he can KMA.
Problem is that AI right now is really bad at catching edge conditions since those people tend to be unique cases without much repetition. However AI is really good at catching ways large numbers of people might have incorrectly calculated taxes similarly. They will try to use it against those big whales but I doubt it will catch many more with only maybe 50x examples to train on, but since it will have millions of examples of poor and middle income people examples it will be very good at catching them.
So long as there is a human being double-checking the calculations. This is vital. We are playing with people's lives here. It is unacceptable for someone to give a cursory look at what AI has spewed out and say "well, if the system says this, it must be correct".
The problem with people at the poorer end of the spectrum is that:-
If they overpay their tax due to AI incompetence: they might have a cash flow crisis which could mean hefty interest and tariff charges which will not be refunded once the anomaly is rectified.
If they underpay their tax due to AI incompetence: it is unlikely* that they will set the balancing amount to one side 'just in case' the tax authority has made a mistake, so when the system comes to them to pay back the overpayment they might well have severe issues paying it back.
*All very well saying that they should, but if you've got a tax employee confirming AI is correct in its calculations, would the tax-payer be inclined to question that belief? I strongly doubt it.