abc.net.au web signal

ABC adopts Anthropic Claude, pilots AI writing with 100 staff

TL;DR

  • ABC has picked Anthropic's Claude as its enterprise-wide AI standard, alongside Microsoft tools and an in-house chatbot called ABC Assist.
  • A July pilot puts 100 "AI Champions" through the tools before staged expansion, with one trial turning regional radio bulletins into digital articles.
  • The ABC will disclose AI use only when it "could materially affect" audience understanding, a narrower stance than its earlier disclosure commitments.

Australia's public broadcaster has just made a concrete AI bet, and the shape of it is more interesting than the usual 'media company piloting a chatbot' headline suggests. According to ABC News, the ABC has picked Anthropic's Claude as its enterprise-wide standard, alongside Microsoft tools and an in-house chatbot called ABC Assist, and will run a July pilot with 100 "AI Champions" before expanding across the organisation.

The most operational piece is a trial that turns regional radio bulletins into digital articles. The same local journalists who produce the radio copy repurpose it into online news, with a local editorial leader and a sub-editor doing a final check before anything is published. Managing director Hugh Marks and chief people officer Deena Amorelli framed the move to staff by saying "the question facing every public broadcaster is not whether they will use AI, but how they will shape the use of AI on their terms and in line with their values."

The politics inside the newsroom is where this gets sharper. Media, Entertainment and Arts Alliance director of media Cassie Derrick said AI used well could free journalists for "the human connection with sources, critical analysis and fact checking," but noted members worry the tools could "compromise job security and audience trust." Recent negotiations won staff consultation rights and ethics guidelines, though management declined to commit that AI would not replace human workers. UNSW professor Deborah Lupton was blunter, arguing the ABC risked "putting their standing at risk" by embracing generative AI at all.

The honest caveat is that the reporting doesn't spell out what triggers the ABC's revised disclosure policy in practice. The broadcaster will now flag AI use only where it "could materially affect their understanding" of content, which is narrower than earlier, broader disclosure commitments. The commercial terms of the Anthropic deal, and the internal metrics that would move the pilot from 100 champions to the whole staff, are not in the story either.

If the regional-bulletin experiment holds up under real editorial pressure, the upside for readers is more written coverage of places that currently only get radio, and Anthropic picks up a very visible public-broadcaster reference customer. The thing worth watching is whether that disclosure threshold ends up doing meaningful work in practice, or quietly reduces labelling overall.

Shared on Bluesky by 2 AI experts