I entered high school right around the time Wikipedia was becoming widely accessible once enough Internet and PC infrastructure had been built out. I remember in my freshman English/Language Arts class, we read John Hersey's _Hiroshima_, and then my class wrote research papers about the use of nuclear weapons and the death counts. We had to be told to not blindly trust Wikipedia, and doing the research showed me how the total excess death counts from the use of bombs was (and still is) disputed with a pretty wide range of estimates.
Now, we can have a chatbot produce an entire paper from either the perspective of "the Japanese were cornered on both the southern front from the US and the north front from the Soviets and the aristocracy were eating rats and they were bound to surrender eventually and so the nuclear use was obviously unnecessary" versus the other perspective which is ubiquitous in the US. Most importantly, the facts regurgitated from the chatbot could cite totally hallucinated sources and made up numbers, numbers potentially poisoned by people putting misleading text on the web for basically free.
Moreover, today, while the death toll estimates in the Gaza Strip since 2023 range from appx 70,000-700,000, in Syria after the fall of Assad in the tens of thousands to over 100,000, the history is being encoded into the training corpus by the side with the most money to spend on producing misleading text.
This isn't speculation, the contracts to do so have been signed and executed. [1] I can't confront the epistemic horror and concern I feel about this.
[1] "Israel Is Paying Millions to Train AI Chatbots How to Talk About Gaza. It's Working." https://www.dropsitenews.com/p/fd107677-6cf9-4ece-9e45-de16bbde8b4a