Almost certainly it does trim the transaction log. (Otherwise the Amazon cloud will fill up with transaction logs.)
Some details from the Amazon site include this topic and two sub-topics that you can investigate: http://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/CHAP_SQLServer.html
When you create a DB instance, you are assigned to the db_owner role for all databases on that instance, and you will have all database-level permissions except for those that are used for backups (Amazon RDS manages backups for you).
For a discussion of the backup model and how you influence it, see:
And, of course, if you change the model from FULL to SIMPLE, the retention period, and so forth that will affect your restore points.
Amazon definitely support point-in-time restores, which means that the transaction logs are being used. But since they do this for all database server types that they support, I do not know the mechanics of what they use to do this.
On the restore process described here: http://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/USER_PIT.html
When you restore a DB instance that is using the SQL Server DB engine to a point in time, each database within that instance will be restored to a point in time within 1 second of each other database within the instance. Transactions that span multiple databases within the instance may be restored inconsistently.
So I would say that, whatever the mechanism of the log backup, it will be available within the constraints discussed in their documentation.