I'll bet it costs a bundle to make sure it works as well as it is politically necessary for it to work. It's a matter of marginal costs and benefits. Train travel is already extremely safe; adding safety measures to an already safe mode of travel is bound to be challenging.
Imagine a world where half the train engineers were stoned out of their mind,and train derailments were an everyday occurrence. It would be cheap to design and install a safety system that would be a huge success by cutting down derailments from a twice a day occurrence to a once-a-month thing. But we live in a world where passenger train derailments, though terrible, are exceedingly rare. They're not even a once-a-year occurrence. This is the first time in a very long time an Amtrak train has derailed for speed. In the past five years the vast majority of Amtrak accidents have been things on the tracks that shouldn't be there or freight trains colliding with Amtrak trains. The last accident a system like the one we're talking about would likely have prevented was in 2011, when an Amtrak train went through a red signal and collided with another Amtrak train.
In our hypothetical scenario if the new system caused one accident a year that'd be a non-concern because of the hundreds of crashes it prevented. But in real life if the system caused just one accident a year that'd represent a tripling of the accident rate ove no system.
You have to have confidence that an automatic system outperforms humans by an order of magnitude before it is accepted by the public, underwriters, investors etc. Otherwise self-driving cars would be a commonplace option already. They already work, probably better than drivers and certainly better than some.