The Harvest: Meta Wanted the Keys to Every Website Behind Its Pages, and Britain's Watchdog Has Finally Looked Up
The UK CMA has opened an investigation into Meta's business terms, which could let it collect publishers' website content for its own products, including AI. Gus Green reports.
Somewhere in a data centre that hums like a cathedral with the lights off, a crawler waits for permission it has already decided to take.
It does not knock.
It does not need to.
On 30 October, the Competition and Markets Authority, Britain's competition watchdog, opened a formal investigation into Meta, the company behind Facebook, Instagram, WhatsApp and Threads.
The question is whether Meta has been abusing a dominant position by imposing what the regulator calls exploitative terms on publishers and advertisers.
At the heart of it is a clause that most people would scroll past on their way to the “accept” button.
Under the new business terms, any company that uses Meta's services could find that it has quietly agreed to let Meta collect content from its own website and use it to build Meta's products.
That includes, in the CMA's words, “its AI features”, and it would happen “without those businesses getting a choice on if or how their content is used.”
Read that twice.
The price of a Facebook Page was about to become the contents of your house.
The Small Print, Grown Large
The story did not begin in Whitehall.
It began, as these things often do, with a journalist reading the terms of service that the rest of us agree to blind.
Nieman Lab, the Harvard journalism outlet, first reported on the new terms on 15 October.
They would allow the scraping of websites “associated with” business accounts on Facebook, Instagram, WhatsApp and Threads.
Virtually every news organisation on earth has at least one of those accounts.
So does the corner bakery, the local football club, the indie bookshop, the plumber, and, for the record, us.
The crawling clause, as Search Engine Journal described it, lets Meta use automated software, “create test credentials” to reach pages, and periodically pull fresh updates.
The Commercial Terms apply to any business use, including simply managing a Facebook Page.
After the outcry, Meta removed the word “train” from the text and narrowed the use of the scraped material to “advertising and business products”.
That edit was quick.
Legal experts who spoke to Nieman Lab remained sceptical that it changed much.
A purpose as broad as “business products” still has room for a very large machine to stretch its legs.
No Door Marked Exit
What makes this feel less like a contract and more like a toll booth is the absence of an off-ramp.
The Professional Publishers Association, which represents British magazine and specialist publishers, put it plainly.
There was no way to opt out.
The only alternative was to stop all business use of Facebook, Instagram, WhatsApp and Threads, which, as the association noted, “is not a realistic option for publishers.”
Meta's reply has been that website owners can still use robots.txt, the small file that tells crawlers where they may and may not go.
That may be true in practice.
The trouble is that the promise does not appear in the terms themselves.
A courtesy you are offered is not a right you hold.
Sajeeda Merali, the association's chief executive, offered the line that should be carved over the door of every platform: “Access to audiences should not come at the price of publishers' content.”
The Movement for an Open Web, a campaign group of advertisers and publishers, filed a formal complaint with the CMA and the European Commission on 23 October, asking regulators to block the terms before they went live.
Its co-founder, Tim Cowen, made the point that haunts every privacy argument of the last twenty years.
“You can't unscrape that content.”
He is right.
Once words, photographs and archives have been swallowed and turned into a model, there is no recall notice, no refund, no way to pull the thread back out of the fabric.
The Shape of the Machine
To understand why this matters beyond newsrooms, it helps to think about what modern artificial intelligence actually eats.
It eats us.
Our sentences, our photographs, our recipes and reviews, our obituaries and match reports, our late-night blog posts about grief and gardening.
The large models that now answer questions, write emails and summarise the news were built on oceans of human writing, much of it taken first and argued about later.
For years, the polite fiction was that the open web was a public park, and anyone could wander in with a notebook.
What Meta's terms proposed was something stranger.
Not a stranger wandering through the park, but the landlord of the town square writing into your lease that he may let himself into your home, take what he likes, and use it to open a shop next door.
The News/Media Alliance in the United States, whose chief executive Danielle Coffey called the terms “outrageous and probably illegal”, asked a question that cuts to the bone.
If Taylor Swift has a Meta business account, does Meta get the right to scrape the media on her website?
If a stock photo library has one, does Meta get its images?
Those are not hypothetical jokes.
They are the logical edge of a clause written to be broad.
What the Watchdog Is, and Is Not, Saying
Here we should slow down, because accuracy is the only weapon that does not rust.
The CMA has not found that Meta broke the law.
It has said, in its own release, that “no decision has yet been made”.
The investigation will first decide whether Meta holds a dominant position in a relevant UK market, and only then whether these terms amount to an abuse of it.
Under the Competition Act 1998, imposing unfair trading conditions can count as abuse, but that is a finding still to come, if it comes at all.
Meta, for its part, told Reuters there were “no grounds” for an investigation.
A spokesperson said the company had “already committed not to change our terms without consulting publishers and other stakeholders” and welcomed “the opportunity to explain our approach”.
The timeline tells its own quiet story.
The terms were due on 30 October.
After what the CMA describes as active engagement, Meta pushed them to 20 November.
On 28 October, it paused the rollout entirely, pending consultation with industry groups.
Two days later, the investigation opened anyway.
For now, the existing terms remain in place.
Sarah Cardell, the CMA's chief executive, said the regulator had heard concerns from news publishers and other businesses and was “acting swiftly”.
In the slow-moving world of competition law, that phrase carries weight.
A Continent Raising Its Hand
Britain is not standing alone in the doorway.
Last week a Norwegian court temporarily banned Meta from enforcing the new terms after ten media outlets argued that they threatened their rights to their own content.
In France, the Paris Court of Appeal rejected Meta's attempt to overturn an order to negotiate in good faith with French media over payment for news used on its networks.
The court found Meta's conduct had caused “serious and immediate harm to the press sector”, and described the data Meta provided in those talks as “fragmentary, opaque and unverifiable”.
Meta said it was disappointed with the French ruling and was considering its next steps.
Since May, more than 340 news sites, including the Guardian, the New York Times and USA Today, have blocked certain archive crawlers to stop their journalism being collected for AI training.
The newsrooms of the world are learning to lock their windows.
The Quiet Theft of Consent
Strip away the competition law and the corporate statements, and what remains is a very old injustice dressed in new code.
It is the injustice of consent extracted rather than given.
Of a choice that only looks like a choice, because the alternative is to vanish from the place where your readers, customers and neighbours actually gather.
A small British publisher cannot simply walk away from Facebook.
A local business cannot pretend Instagram does not exist.
Meta knows this, because the dependence is the product.
We have seen this pattern before.
First the platform invites you in, free of charge, to build your audience on its land.
Then the referral traffic thins, as the publishers' association says its members have seen, while the need to be present stays.
And then, one autumn morning, the small print changes, and the land you built on turns out to have a mineral rights clause.
What is being harvested is not just text.
It is years of reporting by people who stood in courtrooms and floodwater to get it.
It is the photograph a freelancer sold for a modest fee.
It is the review a bakery owner wrote at midnight.
It is labour, and labour taken without payment or permission has an older name than “data”.
The Light in the Server Room
There is a temptation, with stories like this, to shrug.
The terms are paused.
The regulator is on the case.
The machine will be tamed by process.
Perhaps.
But the pause came only after a journalist read the fine print, after trade bodies wrote letters, after campaigners filed complaints and a watchdog started asking questions.
None of that happened by default.
Every one of those steps was a person refusing to scroll past.
That is the lesson worth keeping.
The great harvesters of our age do not usually take by force.
They take by default, by fatigue, by the button we are all too tired to read before pressing.
Consent is no longer something we give.
It is something we fail to withdraw in time.
The CMA's evidence gathering is expected to run into the new year.
Whatever it decides, the question will outlive this case: who owns what we write once a machine has read it?
Somewhere in that humming dark, the crawler is still waiting.
This time, at least, someone has turned on the light.
This is Gus Green, signing off.
And Remember:
Trust no Single Source
Trust Your Gut
and Stay Curious
Gus
For media enquiries, please contact:
UK - 020 3404 2295
USA - 0650 702 7445
AUS - 02 9072 9499
Help Support Gus and the rest of our Authors by visiting the store to get your copy of the latest Anonymous Author Novel or Anything Anonymous.
Visit the store
