The article discusses automatically keeping up-to-date with rapidly changing news broadcasts. Statistical modeling techniques have been successfully applied to the transcription of radio and television broadcasts. This is an exciting research area, in that there are many outstanding issues to be addressed to improve the transcription accuracy on this varied data, and at the same time there are nearterm applications which can be successfully built upon this technology, even though it is imperfect. Radio news shows are usually easier to transcribe than television news shows. The transcription quality is surprisingly good for speech over background music. In contrast, the most difficult portions to transcribe are those containing speech from non-native talkers or overlapping speech such as frequently occurs in interviews or voice-over for translated segments. Keeping the language model and lexicon up to date for breaking news stories is an outstanding challenge for all languages.