The baseband OS is separate from the Linux running on the Raspberry Pi, the modem can be put in hardware RESET state - so, there's unlikely to be a way to execute malicious code on the Linux side by using a modem exploit, if that's what you're interested in =) And yes, the OS has access to GPIOs that control the GSM modem, so that's entirely possible.

Well, I was thinking more that I would want to set up the phone so that the baseband OS is only operational at the time I decide I want to make a phone call. Would something like that be possible and practical?

There's a GSM hardware switch - https://hackaday.io/project/19035/log/60071 - so this is the way to go then, simple as that =) So far, the software is going to assume that the modem is on at all times - however, I'll solve that in the next PCB revision.