Since the release of ChatGPT in 2022, news organizations have published public guidelines to manage the risks of generative AI, but it remains unclear how these documents have changed over time and which skills they implicitly require of journalists. This study examines newsroom AI guidelines across four categories (transparency to audiences, human oversight, accountability for errors, and permitted and prohibited uses) and extracts the skills they imply for journalism education. Using qualitative content analysis, a purposive sample of eight news organizations from the United States, the United Kingdom, Germany and the Philippines was coded with a coding sheet and eight skill tags. Preliminary findings show that transparency, human oversight and accountability appear explicitly in all eight documents. Based on the versions read, one outlet fell into a restrictive level, six into a level of specified uses, and one into a level of generally permitted use under senior approval. The Associated Press (three versions) and the BBC (two) moved from prohibition toward permitted tasks under human review, and AP added a disclosure requirement (2026). Disclosure, verification of machine output, instrumental tool use and human judgment appeared in all eight documents. The guidelines can be read as a map of expected skills; it is proposed that journalism curricula stress verification, disclosure and accountability alongside tool training. The findings reflect stated policy rather than newsroom practice, and the small purposive sample is not generalizable.

generative AI newsroom guidelines journalism journalism education future skills