I was thinking that as well.
If you want to sell manufacturer's machinery or a car or medication, everything needs to go through safety testing. Different fields have different standards. On top of that, different countries have varying legislation which needs to be fulfilled before something can go on the market.
Lots of items are completely forbidden (plastic straws, toys for babys with detachable tiny parts...) or only available in specific settings, such as chemicals or medication.
AI was set into motion with almost no guardrails, and those who are affected have to take measures after the problem already happened - such as all artists whose work was used in training, people in fake videos / posts are just few examples.
Fake material is now so widespread I would class it as guerilla terrorism by the program creators and releasing companies.
When you can't trust the news, and the news media cannot trust anyone of their contributors, people can unleash heaps of trouble on anyone without reason by generating fake material.
Sure, there was fraud and fake videos or photos before, but the difference is the sheer relentless amount and that it is open to anyone.
Depending on the training material, there is also a lot of inbuilt bias.
Our politicians were incredibly shortsighed and not that bright to permit this software onto the market without any regulation. Yes, the EU is talking about a few things here and there, but it's late and not coherent. The problem is already there.
Additionally, the promise was that AI would help in workplaces which include heavy manual labor, are tough on the body, such as patient care (lifting patients, for example), tough mining work, waste collection, and similar, which also have problems with hiring workforce. Instead AI starts cutting into the office and creative work first.
There should be more focus on using AI or automation in areas where there is a problem retaining workforce or the work is dangerous or often leads to health problems. This can be done by legislation and agreements with the tech companies.
It's also stupid to de-skill oneself. Sure, ChatGPD can do your homework, but you won't have learned anything.
Regarding the environmental impact, Meta & Co are planning mini nuclear reactors to power their servers, so much energy is required.