Agreed, since these tech companies are mainly leftist/woke, they ultimately will have the desire to control or limit what the AI can output so that it works for their agendas.

AI's are mathematically objective in what they learn, so the only real ways to get them to be "woke" is either by limiting the training data, or to add in filters before and/or after the neural net, or trimming out the segments of the net that lead to "problematic" outputs... but that's tricky because if you have a matrix of 0.2343, 0.8495, 0.3323, to the n, it's hard to know which neurons are the root.