AI mistakes in newsrooms
Social posts this week flagged visible errors from AI‑generated journalism — one example quoted a mangled Dave Grohl line as “fish try to avoid the skies.” (x.com) Separately, Cambridge PhD Raphael Hernandes was nominated for a European Press Prize for work that used AI to analyse far‑right Facebook rhetoric, showing both innovation and controversy in the same moment. (x.com)
Newsrooms are getting two public lessons in AI at once: visible publishing mistakes can spread fast, and careful AI analysis can still win major journalism recognition. (ap.org) One example circulating on social media this week showed an AI-generated article mangling a Dave Grohl lyric into “fish try to avoid the skies,” a line users shared as evidence that automated copy can fail in ways human editors would usually catch. (x.com) At the same time, Cambridge Digital Humanities said on April 16 that PhD student Raphael Hernandes was shortlisted for the 2026 European Press Prize Innovation Award for a Guardian investigation into far-right Facebook groups. (cdh.cam.ac.uk) Cambridge said Hernandes and his colleagues spent a year investigating links between the 2024 United Kingdom summer riots and far-right Facebook groups, using network analysis and large language model labeling to examine 51,000 social media posts. (cdh.cam.ac.uk) The European Press Prize shortlist names the Guardian project, “Inside the everyday Facebook networks where far-right ideas grow,” as a 2026 Innovation Award nominee alongside four other projects. (europeanpressprize.com) That split screen reflects how AI is actually entering journalism: as a text generator that can invent or garble facts, and as a data tool that can help reporters sort large volumes of material for patterns. The Associated Press says any output from a generative AI tool should be treated as “unvetted source material.” (ap.org) The Associated Press also says it does not use ChatGPT to create publishable content, even though staff may experiment with it cautiously. In a later update, AP said it would allow limited uses including Spanish translations, article summaries, and headline suggestions, with an AP journalist editing and vetting the work before publication. (ap.org 1) (ap.org 2) A survey released by AP in 2024 found nearly 300 journalists and newsroom leaders had already seen generative AI reshape workflows, and 70% said their organization had used it in some capacity. Almost half said tasks or workflows had already changed because of it. (ap.org) Public rules still lag behind adoption. The American Journalism Project said in November 2025 that only about 20% of local news organizations had public AI usage policies, citing a worldwide study of 101 journalists, even as more outlets folded AI into editorial work. (theajp.org) Hernandes described his shortlisted work as an inquiry into “a network of UK-based Facebook groups acting as radicalization engines for far-right rhetoric,” and Cambridge said his current research examines how AI reshapes journalism and information environments. (cam.ac.uk) (europeanpressprize.com) So the current argument in newsrooms is not whether AI is present. It is whether editors can keep synthetic errors out of published copy while using machine-assisted methods on datasets large enough that reporters cannot read every line by hand. (ap.org) (cdh.cam.ac.uk)
