Anime

You're conflating two entirely different debates.

Yes - there has been some debate around western publishers overly aggressively "localizing" manga and/or changing details to not just make things more understandable to western readers, but deliberately altering social /political content to accord with their own views. The two broad positions in that debate are to continue to depend on western publishers and their translations, or to keep translation in-house - under the supervision of the Japanese publishers.

This debate starts from the position that translation will be kept in-house, and concerns how it will be done - whether by human translators or AI. The publishers want to use AI for one and only one reason - because it would be cheaper. The JAT's position is that machine translation is so vastly inferior that it will not work, and that human translators must be used.

The AI translation argument wasn't really about whether nor not it would work, it was people voicing their frustration at Western Localizers making large alterations to apply their own personal culture and politics to a work that did not originally have that.

That was not the argument at all. Japanese publishers want this to be able to increase production overseas. That was the argument. Translators don't want this because a: it's their jobs, and b: it's very inaccurate.

https://animehunch.com/japanese-govt-major-manga-publisher-invest-heavily-in-ai-translation-to-boost-manga-export-overseas/

Also, the politics insertion is on such a small scale that it doesn't warrant full scale AI translation in any way. Many of these cases are due to editors anyway and not the translators so AI would not fix this problem. Localization itself is very much necessary. Japanese is conceptually such a different language system to English.

And honestly, I would prefer grammatically inaccurate machine translations over "localized" translations that deviate significantly from the original.

These 'grammatically inaccurate machine translations' often are so bad that they don't even get the subject of a sentence correct, completely changing the meaning. Or don't have context and as such, writes something that completely doesn't relate to what was previously written. Calling it just grammatically inaccurate is downplaying how bad it can be in my opinion. Human QA can remedy some of it so that it's more readable, but that's not enough to actually fix bad flow, misinterpreted meaning, missing context etc. Having to go back and fix shit is much more intensive than doing it right in the first place.