Most of the things we read online about AI is just advertisement by the companies trying to appear "useful" to solve complex problems nobody was able to solve. We may even call it a scam to some extent. The so called "mathematical problems" they claim to have solved aren't as cool as they claim.
Take the Navier–Stokes fluid dynamic that OpenAI claimed to have solved. First of all it is a big leap to claim they can also solve ECDLP even if they had actually solved it. But also they didn't actually solve anything. It was just publicity stunt. Their tool used an existing model developed in 2023 to get closer and the Clay Mathematics Institute still categorizes it as "active" not solved.
And yet in the media they are still claiming AI has solved this!!!
As for the usual quantum computers, we still need computation power to solve ECDLP and for the coins to be at risk. AI is not going to change that.
That make sense, just another stunt to inflate their IPO, never really looked into all these math discoveries and what each has discovered, whether it's truly a discovery or just nothing.
It is highly unfortunate that it is people that will be moving funds to a new address when the wallet can do it automatically, but altcoins wallet developers make it not possible just as many of them do not make it possible for bitcoin. But bitcoin wallets that are open source and reputed will automatically send change to change address. What that matter is that you should not reuse the same address to receive coins the second time, just use it only ones.
ETH wallet makers are more focused on in-app swap feature that charges 0.8% a swap on top of the pool fee instead of features like this, won't be surprised at all if they are simply telling us to generate new seed phrase

.
If there were an imminent threat regarding the weakness of entropy in Bitcoin and popular wallets, these measures would have been urgent. I see this research as an indirect promotion of what OpenAI can do and thus obtain more funding and loans.
Could be one of the reasons indeed, but I suppose entropy in wallets as long as it's enough, should be safe because no algorithm can just bypass physic due to pure randomness. The problem is always on the public key.