Associated Press issues newsroom guidelines on generative AI use

The Associated Press issued formal newsroom guidelines governing staff use of generative AI tools such as ChatGPT. The rules barred using AI to create content for publication and barred altering photos, video or audio with generative tools; they also prohibited publishing AI-generated images suspected of depicting false events, unless clearly labelled as illustration and relevant to a story.

Limited uses were permitted: staff could use AI to draft internal summaries, brainstorm headlines, or assist with tasks like transcription, but any such output was to be treated, in the wording reported by Poynter, as “unvetted source material” requiring the same editorial judgement and verification as a human tip. AP’s vice president for standards, Amanda Barrett, said the tools were useful but “does not replace the journalistic smarts, experience, expertise and ability to do our jobs.” The guidance also flagged AI’s tendency to fabricate information, directing staff to apply ordinary verification practices, including reverse image searches, before relying on any AI output.

AP was among the first major wire services — whose copy and style conventions are followed by thousands of smaller newsrooms — to formalise a generative-AI policy, and the guidelines became a reference point against which other outlets’ subsequent policies were compared. The move followed AP’s earlier, narrower 2023 partnership allowing licensed use of its text archive to train OpenAI’s models, and came two weeks before a widely mocked incident at Gannett underscored the risks the guidelines were meant to pre-empt.