The current online version of the Java Language Specification, section 3.1 (http://java.sun.com/docs/books/jls/html/3.doc.html#95413), states that "Java programs are written using the Unicode character set, version 2.0." It then goes on to explain that older versions of Java used older versions of Unicode.

In fact, since JDK 1.1.7 we've been based on Unicode 2.1, and we're planning to upgrade to Unicode 3.0 in the Merlin release. It's probably not a good idea to burden the language specification with tracking these details. We should instead write something to the effect that Java tracks the Unicode standard and that in future the Unicode version for each Java version will be specified in the javadoc for the Character class (there is a bug 4247314 to that effect). Then, since older versions of the Character class had incorrect or no information on the subject, historical information might be given:
JDK 1.0.x -> Unicode 1.1.5
JDK 1.1 to JDK 1.1.6 -> Unicode 2.0
JDK 1.1.7 and up, Java 2 SE 1.2.x and Java 2 SE 1.3.x -> Unicode 2.1