The Bearing isn’t the only place experimenting with AI news.

The ABC has signed a deal with Anthropic to bring Claude into its newsrooms, and is separately piloting a tool that converts regional radio bulletins into written articles. About 100 staff are involved at this stage, with the scope likely to expand. The goal is to free journalists from routine production tasks so they can do the work that actually requires a journalist: investigations, source relationships, editorial judgement. That is a sensible ambition. The editorial standards governing it remain largely undefined.

The efficiency case is real — the question is where the gains go

The case for this kind of adoption is real. Journalists have been using automated systems to convert data into routine news copy for over a decade, and a 2016 study found readers actually rated machine-written articles as more credible and higher in journalistic expertise than human-written equivalents, which says something about the floor of competence at the bottom of the production pile. More ambitiously, the BBC used AI to process enormous volumes of text and video from Ukraine in ways that simply would not have been achievable manually. News organisations in the Global South have used it to translate and repurpose content across language barriers with minimal resources. The efficiency gains are not theoretical.

The risk is specific to the ABC's circumstances, not to AI adoption in general. The broadcaster is not operating from a position of public confidence. Australians already carry a distinctive distrust of AI tools, and the ABC has spent years navigating accusations of bias that, whatever their merit, have left a portion of its audience primed to notice any misstep. A hallucinated detail in a ChatGPT-assisted legal brief is embarrassing. The same error in a publicly funded news bulletin, produced with a tool the audience already suspects, is a different kind of problem. The profession is facing a crisis of public trust and AI introduces a new vector for errors that can be plausible-sounding and hard to catch.

The question is whether newsrooms under resource pressure will actually spend that freed-up time on verification, or whether efficiency gains will be captured elsewhere, through fewer staff, more output, or both.

The counter to that risk is verification, and verification is precisely what the ABC says it wants to protect. The logic of using AI for routine production is that it buys back time for the editorial work that actually requires human judgement. If that trade holds in practice, the model is defensible. Journalists become curators and checkers rather than transcribers, which is arguably where their value has always been highest. The question is whether newsrooms under resource pressure will actually spend that freed-up time on verification, or whether efficiency gains will be captured elsewhere, through fewer staff, more output, or both.

AI adoption in a shrinking media economy cuts both ways

That pressure is structural. The business models that sustained Australian journalism have been gutting themselves for fifteen years. As we've covered previously, the ABC's own advertising blitz for its own brand reflects a funding model under slow-motion strain, not a communications strategy. AI adoption in that context carries a dual character: it is both a genuine tool for doing more with less, and a mechanism that could make it easier to justify doing journalism with fewer journalists.

There is also a competition problem that sits upstream of any editorial question. AI-generated summaries served through search engines are already diverting traffic away from original news sources. Australian outlets, including the ABC, are competing not just with each other but with an effectively unlimited international content supply. If the ABC's AI-assisted output eventually gets subsumed into the same AI-generated media environment it is trying to navigate, the efficiency gains become circular. You use AI to produce more content, so that AI can summarise it for an audience that never visits your site.

None of this makes the ABC's trial wrong. The efficiency argument is legitimate, the historical precedent for journalists adopting new production tools is strong, and the specific use cases being piloted, converting radio to text, processing large datasets, freeing time for investigations, are exactly the places where the risk-to-benefit ratio looks favourable. But the framework governing what gets checked, how errors are caught, and who is accountable when something goes wrong has not been made public. For a broadcaster funded by taxpayers and accountable to a public that already views it with some suspicion, that transparency is not optional. The tool is fine. The editorial standards around it still need to be shown.

Sources

The Conversation — ABC will trial using AI for journalism. What are the risks and benefits?

The Bearing — Why the ABC is paying to advertise itself — and what that tells us about public broadcasting

The Bearing — Labor sets AI rules before the window closes — but offers few details on what they'll actually do

Frequently Asked Questions

What AI tools is the ABC using in its newsrooms?
The ABC has signed a deal with Anthropic to use Claude and is separately piloting a tool that converts regional radio bulletins into written articles. About 100 staff are involved in the current trial, with the scope expected to expand.

Why does AI in journalism pose a particular risk for the ABC?
The ABC is a publicly funded broadcaster that already faces accusations of bias from a portion of its audience. An AI-generated error in a news bulletin — plausible-sounding and hard to catch — carries reputational consequences that are more serious for a taxpayer-funded outlet under scrutiny than for a commercial competitor.

Can AI actually improve journalism quality, or does it just cut costs?
The case for improvement rests on a specific trade: AI handles routine production tasks, freeing journalists to spend more time on verification, investigations, and source relationships. Whether newsrooms under financial pressure will use that freed time that way — rather than reduce headcount — is the central unresolved question.

What happens to news organisations that use AI to produce content when search engines also use AI to summarise it?
AI-generated summaries in search results divert readers away from the original source, reducing the traffic that publishers depend on. This creates a circular problem: outlets use AI to produce more content, but AI summaries mean fewer readers ever visit the site to consume it.

Have news organisations used AI successfully before?
Yes. Automated systems have converted data into routine news copy for over a decade, and the BBC used AI to process large volumes of text and video from Ukraine at a scale that would have been impossible manually. News organisations in the Global South have also used it to translate and repurpose content across language barriers with minimal resources.