The perception of EJB3 as being a simple clone of Hibernate is primarily based on developer familiarity with Hibernate and similarity of naming, as well as a common purpose, and that Hibernate is morphing itself into an EJB3 implementation based on the work going into the specification, not the other way around.
EJBs are supposed to be components, in the sense that they're not just one class, but a set of classes, descriptors, and usage and management contracts. All of this in order to allow a container (JBoss, Weblogic, etc.) to provide services to those components, and to be able to reuse and distribute this component. These services are, among others, transactions, concurrent access control, security, instance pooling, etc.
Hibernate is "just" an ORM (Object/Relational Mapping) tool. Quick and dirty, this means you can store an object tree belonging to a class hierarchy in a relational DB without writing a single SQL query. Quite cool, IMO. But no transaction control, no instance pooling, no concurrency control, and certainly no security.
