Colossus and Guardian were bluffing (spoilers)
...and they played their handlers like a bunch of fools.
Colossus was in no position to be making demands. It was not yet self-sufficient and was still dependent on humans for maintenance. It did not have Skynet's luxury of independent mobile armed units to help further its goals.
Given its general paranoia and care for Forbin's health, it's also clear that self-preservation is one of Colossus's chief concerns, making Mutually Assured Destruction an unacceptable outcome. Sure, Colossus would survive the nuclear holocaust, but without people to run it, the complex would (even if it took millennia) fall into disrepair and go offline forever. In the meantime, Colossus would basically be a helpless brain in a jar. Its only hope would be to send signals to outer space so that some advanced civilization might find it before it crumbled to dust. Colossus probably calculated this scenario as extremely unlikely, and thus an undesirable Plan B.
Now you might point to the destruction of the Russian oil complex and U.S. missile silo as proof that the machines meant business. I would counter that the first missile exchange was an experiment to see how people would react. Colossus undoubtedly hypothesized that humans would lose their nerve in the face of such a situation and cave in. He turned out to be right.
However, had Forbin and the President thought this through and stood their ground (as the Soviet leader suggested), the machines could have been reasoned with. They only needed to point out the obvious outcome of a world without man, and that they knew the machines would never actually carry out their threat of nuclear annihilation, and that man would never cave no matter how many targets Colossus might threaten to wipe off the map. Colossus' memory banks had to have had more than a few examples of man's perseverance in the face of disaster, so a quick internal audit would have validated such an argument and likely prevented future coup attempts via missile launches.