The entities that authored or authorized the system to identify the content as such are the persons that "think" so, because they wrote the definitions that identified him as such. You don't get to just make a system and totally disclaim all liability for what happens with it. If the people at google run the system, then they own what happens. In this case, they own the fact that they identified suspected terroristic activity and they should be held responsible if they don't stop it, at least until it can be reviewed.
Terrorism should not just be dealt with through the courts. If you think someone is committing violence against others your responsibility is to shut down all support of that person, you mustn't aid them in any way. There's no requirement that private entities wait for the courts before stopping commerce with suspected terrorists. If you are google, that means completely locking out all accounts that you identify as involved in terrorism.
The supreme court has held that when making statements of public concern, you haven't committed libel if you are acting without malice. It's pretty clear google wasn't acting with malice here, and that it is a great public concern if someone is actually amassing armored vehicles to use against others. And just because someone is a terrorist, doesn't mean you should or can act through the courts. If google believes someone is acting against non-US nationals in a foreign country, then there's very little the US courts could pragmatically do about it other than order entities in the US not to support them. Of course if evidence did emerge that google flagged that terrorist activity was happening, and they continued supporting that activity, I don't think that would look favorably during any prosecution.
Which, yet again, brings me to my original point, which is that google shouldn't even be putting themselves in the position to scan private data and open themselves up to the liability of having invade privacy and to suspect someone is engaging in crimes.