I ran into a interesting article that sheds some light on this...
A number of early users of the application report that Apple is tapping Session Initiation Protocol (SIP) as the basis for launching streaming audio or video. To invite someone to a multimedia session, iChat AV appears to send a request via AIM. The recipient's iChat AV client receives the request and automatically opens the appropriate ports to receive an incoming SIP INVITE message.
Shortly thereafter, the sender's client is able to send a SIP INVITE to the recipient's IP address, which has been provided by AIM -- in this way, the participants' IP addresses have been exchanged invisibly. After using SIP to set up a multimedia session, the actual video is handled via RTP, the real-time transport protocol.
While it uses SIP to launch a multimedia session, iChat AV doesn't seem to function as a full-featured SIP client, however. Aside from receiving initiation requests in conjunction with AIM, the client can't initiate Voice-over-IP sessions using standard SIP INVITE messages --
for instance, those originating from a SIP phone.