I've had something similar happen to me but I can understand the reaction of the AI. They can only see transports moving close to their planets. The fact you're at war with someone else already doesn't mean you can't or won't declare war on them. Destinations of fleets can be changed at a moment's notice so the AI can't rely on that information too much either.

In short, I don't know how an AI could positively confirm you're not going to attack him but some other hapless victim that just happens to be near him. Given the developer's reaction to the often-used surprise attacks players used to knock out the AI (just see how they always vote 'yes' on the no-surprise-attack motion) it's understandable that they make sure the AI pays close attention to it. In the mind of the AI, wouldn't it be better to err on the safe side? It knows us humans just can't be trusted (despite our puppy eyes).