Well, they say they are. But the nondisparagement agreement repeatedly forbids revealing the agreement itself, so if it wasn't cancelled those subject to it would be forbidden to point out that the public claim they were going to release people from it was a lie (or done only for people from whom OpenAI was not particularly concerned about potential disparagement.)
“disparagement" is whatever is defined in the agreement, which reportedly (from one of the people who declined to sign it) includes discussing the existence of the agreement.
