Can we trust their developers to protect our privacy and security?

There is little hope of that if even article writers can’t be trusted to ask the correct questions such as “Do the AI model developers have enough control over their models to even tell if the model would do/not do that?”