Just like TCP, it’s a stateful protocol.
SYN
SYN-ACK
ACKHe might also just see that B is making inputs and assume, if they are being lax, but I assume they wouldn't.
Imagine being in an emergency, possibly listening to someone speaking while wearing an oxygen mask, in a smoke filled cabin, with alarms going off, and asking yourself - "wait did they say 'you ^have^ the controls', or 'you _have_ the controls'".
Err… isn’t that exactly how Air France Flight 447 crashed? Both people thought they were flying, doing opposite things, the plane cancelled out their inputs and did nothing.