The Australian Broadcasting Corporation wants you to believe that artificial intelligence companies are modern-day cattle rustlers, sneaking across digital borders under cover of darkness to swipe sacred journalistic livestock. When Anthropic recently floated an opt-out model for Australian publishers, the mainstream press reacted with predictable pearl-clutching. ABC executives warned of imminent media cannibalization. Trade associations screamed about fair compensation. Everyone involved is looking at the board through a rear-view mirror.
Let us be entirely clear about what is actually happening. Traditional media outlets are not panicking because machine learning models are stealing their intellectual property. They are panicking because these systems are systematically dismantling the economic rent-seeking that has sustained legacy media for the past three decades. For a different view, consider: this related article.
I have watched traditional publishing boardrooms burn through millions of dollars chasing platform distribution deals, algorithm tweaks, and paywall workarounds while ignoring the fundamental structural shift in how humans consume information. Screaming about copyright infringement is easier than admitting that your distribution channel was completely dependent on search engines and social media platforms acting as tollbooths.
The opt-out debate is a red herring. By focusing on whether publishers should be allowed to check a box that blocks web crawlers, the legacy press is successfully distracting everyone from the real crisis. The issue is not that artificial intelligence is reading your news. The issue is that nobody needs to read your bloated, ad-cluttered three-thousand-word feature when a language model can synthesize the underlying facts in two seconds. Similar reporting on the subject has been published by Mashable.
The Myth of the Free-Rider Problem
The entire legislative framework governing digital content in Australia operates on a flawed premise. Lawmakers bought into the fiction that training a machine learning model on publicly available web text is functionally identical to photocopying a textbook and selling it on a street corner.
This is economically illiterate.
When an algorithm parses an article about a Senate inquiry, a corporate merger, or a local council vote, it is not retaining a copyrighted work in a digital filing cabinet. It is extracting statistical relationships between words. It is learning grammar, syntax, and factual correlations. Copyright law was specifically designed to prevent market substitution—preventing someone from selling a competing copy of your exact work. Summarizing public facts and synthesizing general knowledge does not constitute market substitution. If it did, human journalists would be paying royalties to every historical newspaper archive they browsed at the local library.
When publishers demand mandatory licensing fees for machine learning training data, they are essentially attempting to patent the English language and historical facts. If you report that a bridge collapsed in Sydney at 8 AM, you own the specific prose of your article. You do not own the fact that the bridge collapsed. Yet the current regulatory push treats every single data point on the internet as proprietary private property.
Australia tried this road before. The News Media Bargaining Code forced tech giants to pay local publishers massive subsidies, celebrated locally as a great victory for journalism. What was the actual result? The money flowed almost exclusively to legacy conglomerates, propping up bloated corporate structures while independent, local investigative outlets continued to wither. It created a cartel of incumbents protected by government mandate, entirely insulated from market realities.
What Cannibalization Actually Looks Like
Let us address the ABC's favorite buzzword: cannibalization.
Media executives love to claim that generative models are devouring their audience by answering user queries directly in chat interfaces, starving websites of referral traffic. They are right, but they are drawing the wrong conclusion.
The referral traffic model was always a toxic bargain. For twenty years, publishers traded direct reader relationships for algorithmic handouts from search and social platforms. They optimized headlines for click-through rates, buried reporting beneath walls of autoplay video ads, and bloated page loads with third-party tracking scripts that would make a privacy regulator weep. Readers hated it. Users did not visit news sites because they loved the user experience; they visited because they had no alternative bottleneck to reach the information.
Now, that bottleneck is gone. When an interface gives users a direct, ad-free synthesis of breaking news, traffic drops. Of course it drops. Why navigate a minefield of banner ads and cookie consent pop-ups just to read three paragraphs of actual reporting buried beneath nineteen paragraphs of search-engine-optimized keyword repetition?
The migration away from traditional web browsing toward conversational information retrieval is not a theft of value. It is a market correction. The legacy web business model relied on artificially inflating friction to maximize ad impressions. Artificial intelligence removes that friction. If your entire business model depends on forcing human beings to wade through digital garbage to find a single fact, your business model deserves to die.
The Opt-Out Trap
Anthropic suggesting an opt-out mechanism sounds polite and accommodating, but it is a trap for the unwary. If a publisher chooses to opt out—blocking crawlers from indexing their site—what happens?
They disappear from the training data of the next generation of intelligence models. In the short term, they might feel righteous protecting their precious paragraphs. In the medium term, they render themselves invisible to the most important distribution layer emerging in modern technology.
Imagine refusing to let your business be listed in a digital directory because you want people to walk past your physical storefront instead. It is commercial self-harm disguised as a principled stand.
Smaller, nimble publications understand this. They are leaning into direct reader relationships, deep-domain authority, and original investigative reporting that cannot be easily synthesized from public wire copy. They are not wasting legal budgets lobbying for government protection against math. They are adapting to a world where attention is scarce and commodity reporting has zero marginal value.
The Uncomfortable Truth About Journalism
Here is the part nobody in a legacy newsroom wants to say out loud: ninety percent of daily news content is pure commodity.
Press releases rewritten with a different headline. Police blotter summaries. Financial earnings reports churned out by templates. Syndicate wire copy duplicated across fifty different mastheads with minor stylistic tweaks. When an artificial intelligence model ingests this material, it is processing noise, not signal.
The panic over training data assumes that every piece of published journalism is a unique work of high art. It is not. It is industrial-scale content production designed to feed the programmatic advertising beast.
Real investigative reporting—the kind that takes months of source cultivation, freedom of information requests, and legal risk—is expensive, rare, and profoundly valuable. And guess what? High-end investigative journalism is largely immune to automation cannibalization because it relies on discovering new facts, not rearranging existing ones. If you break a massive corruption scandal, every artificial intelligence model will cite your reporting as the primary source, driving massive credibility and direct authority back to your brand.
The outlets screaming the loudest about copyright theft are rarely the hard-hitting investigative outfits. They are the content mills realizing that their low-effort SEO aggregation strategy has an expiration date.
Where the Market Goes From Here
If you run a media company today, stop spending your capital on lobbying parliaments for protectionist rent-extraction schemes. It is a dead-end strategy that buys you time while destroying your organizational agility.
Do three things immediately:
First, audit your inventory. If your traffic relies entirely on search engine optimization and commodity news aggregation, accept that your current business is a melting ice cube. No amount of licensing revenue from technology startups will save a product nobody actually wants to use.
Second, pivot to un-automatable value. Build direct subscriber communities, proprietary data analysis, and domain-specific vertical reporting that requires deep human expertise. If your output can be fully replicated by a competent intern reading a wire service, an algorithm can do it in a millisecond.
Third, stop treating distribution platforms as enemies. The winners of the next decade will not be the companies that successfully lock their content behind impenetrable paywalls and legal threats. They will be the organizations that integrate fluidly into new mediums, establishing their brand as the undeniable ground truth that models rely on and cite.
The ghosts of print media past spent twenty years fighting the internet. Do not spend the next twenty years fighting mathematics.