Meta’s re-entry into media partnerships, on the surface, might sound like a win for only a handful of publishers. After all, haven’t we seen this pick-and-choose behaviour before?

This, however, is different. 

Meta’s recent spate of media deal signings is part of its strategy to fix its AI model. And it is Meta’s pivot acknowledging it has to pay for training content that will prove to be the much bigger deal for the wider tech and media worlds.

To get up to speed, let’s unpack what’s happened with Meta’s love-hate-love relationship with news media. We’ll also explain why News Corp’s new mega-deal excludes Australia, despite its world-leading media payment laws. And why Canada, too, is still missing out on the Meta love.

1. What Meta is doing 

Since December, Meta has announced a limited number of media partnerships, capped by a US$50 million a year deal with News Corp for U.S. and U.K. content over three years.

2. Its deal book

Meta has also signed deals with: USA Today Co., People Inc., CNN, Fox News, Fox Sports, The Daily Caller, Washington Examiner in the United States, and Le Monde in France. In March 2026, it added deals in Spain, Germany, and France: Grupo Prisa, Süddeutsche Zeitung, and Le Figaro. These followed Meta’s signing with Reuters to use its news content in its chatbot. 

3. Why Meta says it’s doing it

“We’re committed to making Meta AI more responsive, accurate, and balanced,” the company said when announcing its first AI deals. Meta said it added licensed news sources because AI systems struggled to stay current on breaking events. 

“When you ask Meta about news-related questions, you’ll now receive information and links that draw from more diverse content sources to help you discover timely and relevant content tailored to your interests.” 

4. Why Meta is actually doing it

Meta has fallen well behind with its own foundational AI model, Llama, and needs to feast on reliable content to catch up. Most immediately, it needs a top-rate answer engine to sustain its advertising powerhouse. 

AI investor Gavin Baker, the founder and chief investment officer at Atreides Management, said on Patrick O’Shaughnessy’s podcast that Meta was probably not in the top 100 models, let alone a frontier model. This despite Mark Zuckerberg predicting at the start of last year that Llama would become No. 1. Baker said Zuckerberg was “as wrong as it was possible to be.”

5. The bigger picture

You need to look at how Meta built Llama to understand its woes.

It began as an “open source” model that was neither open source, according to the independent body that governs such things, nor very good. So, developers and users failed to flock to it, which deprived it of content, which slowed its progress further. 

Meta relied on user-generated content from Facebook and Instagram and became a big user of synthetic, AI-generated, training data. This also failed to move the dial. 

Unlike Google — which supplements its avalanche of user-generated Reddit with journalism deals — Meta has refused to pay media companies for any training content. 

6. How they are addressing the problem

Meta eventually realised it has to pay to get robust content to feed its LLM. It began by signing licences with a handful of publishers to provide credible content for both model training and real-time data retrieval. Specifically, that means news. 

Meta started with a limited number of contracts in the United States and one or two in smaller language markets, much as OpenAI did to build just enough coverage for reliable, topical answers. Meta has also thrown hundreds of millions of dollars at attracting talent in the AI research race to lead its AI reset but has lost as many big names as it’s gained. 

Until it stabilised, Meta delayed the rollout of its replacement for Llama 4, dubbed Muse Spark, which becomes the first release from the new Meta Superintelligence Labs. It has now abandoned its open-source approach, soft-launching in April 2026 to selected Facebook and Instagram customers. 

7. How News Corp earned the gold medal

News Corp sits at the top of the table for commercial deals with tech platforms. The company is set to earn US$50 million a year from OpenAI for five years, US$50 million a year from Meta for at least three years, and is widely believed to earn much more again than those deals from its long-running global Google partnership — although that partnership does not yet include any AI training component. 

News Corp also has a long-standing arrangement for its content to appear in the Apple News+ subscription service. 

News Corp’s U.S., U.K., and Australian businesses have given it unmatched negotiating experience, with monetary and data sharing benchmarks. It focuses on licensing deals, not revenue share, and where possible deals must be global. 

Its sustained, high-profile campaign against Big Tech over content payments has paid off. News Corp chief executive Robert Thomson has prosecuted this in public, more loudly than any other leader, for nearly a decade, initially on anti-trust grounds and, in the AI era, through a simplified woo-or-sue doctrine. 

His March 2026 version: “We’ll woo you, we’d like you to be our partner. But if you’re stealing our stuff, we are going to sue you.” 

All eyes are on whether it will pull off the same sized deals with Anthropic, TikTok, and Microsoft as the copyright payment gates are pried open. 

Meanwhile, News Corp’s legal case against Perplexity continues in court.

8. United Kingdom included, Australia excluded

At the peak of Facebook’s media deals, Meta struck a large three-year commercial agreement with News Corp in 2021, covering its Australian brands after the country’s media bargaining law passed. It already had a deal with News Corp in the United States, but its U.K. titles, including The Times and The Sun, were left out. Meta was waiting to see the terms in the U.K. government’s media bill, which has never materialised.

In the latest deal, News UK is in, but News Corp Australia is out. This time Meta is waiting to see the final terms of Australia’s update to its media bargaining law, in particular to see the shape of promised incentives. Meta’s earlier contracts have expired, public consultation on the new law has closed, and the release of the government’s final response has dragged on despite bipartisan political support. It’s now expected to be released in the next month. U.S.-Australia trade talks have also played a role in the delay.

While Meta is expected to return to the table as soon as Australia’s legal update passes — which for News Corp will be paid on top of the US$50 million already announced — there is a political complication. The new Australian proposal would capture more tech companies in its net, but it does not address AI training. 

Even if Meta were to put an offer on the table today, News Corp would be unlikely to sign on before the legislation goes to a vote in Parliament. To do so would risk easing pressure on the Australian government to pass a comprehensive overhaul. What News Corp says and does matters enormously in its home market. The original bargaining code, which shaped this debate globally, in a real sense only existed because of News Corp’s relentless campaign.

9. Canada sidelined

Meta’s ban on news in Canada remains in force. Talks with the government resumed in January 2026 to find a path through Meta’s objections to the Online News Act but with no breakthrough. The government says it remains open to striking a peace deal. 

Canada’s law requires dominant platforms to compensate publishers for news content made available on their services — a more refined version of the original Australian law. (In Australia, Meta had also switched off news but backed down after three days and a torrent of bad press.) 

More than two years after Meta began blocking all news on Facebook and Instagram in Canada, users still cannot view or share news content. And publishers still cannot use those platforms to distribute any form of journalism to audiences in Canada. 

Google agreed to comply with the law via a C$100 million annual payment to media companies that it negotiated directly — and highly unusually — with the government. Meta has held firm in Canada even as its own corporate policies towards media deals elsewhere have rapidly softened.   

10. A love-hate-love of news

Meta’s relationship with news publishers is embodied in Cory Doctorow’s Enshittification theory. It describes how tech giants grow by befriending users and partners, then by betraying them progressively as they ascend to market dominance. 

In 2024, when it shut the Facebook Journalism Project and News tab, Meta argued people did not come to Facebook for news and that news created little value for the company. “News makes up less than 3% of what people around the world see in their Facebook feed,” it said. 

This ignores the central role news content plays in its users lives, with studies consistently showing most people get news from social media. Meta abruptly exited all of its partnerships and as an exclamation point initiated the news block in Canada. 

What the Entshittification bell curve lacks, however, is our current phase: a possible relationship on the rebound as new partners are signed. 

An early signal of Meta’s renewed engagement with the journalism sector can also be found within the social giant’s recent tweaks to its news content algorithms. Referral traffic to publishers from Facebook and Instagram is growing in many markets. Chartbeat has recorded modest lifts tied to AI-driven discovery and linking. It is tentative, uneven, and it won’t result in a return to the bygone traffic era, but it’s a double-digital turnaround for some publishing houses. 

11. Who’s suing Meta

Legal pressure on AI companies about their model training continues before the courts, but Meta is the defendant in only a handful of the more than 50 notable cases. Authors, publishers, artists, and other rights holders are pursuing cases over scraping, copyright infringement, and the use of protected material in training data. The outcomes are still uncertain. 

Yet the trend is unmistakable as the value of content is contested not only in commercial negotiations but in courtrooms as well. 

Here are the current IP cases involving Meta:

• French publishers and authors vs. Meta: French publishing and authors’ groups sued Meta in Paris over alleged large-scale use of protected works for AI training.

• Entrepreneur Media vs. Meta: Entrepreneur’s publisher alleges Meta used copyrighted magazine, online, and book content to train Llama without authorisation.

• Kadrey et al vs. Meta: Authors including Richard Kadrey, Sarah Silverman, Ta-Nehisi Coates, and Junot Díaz allege Meta used pirated books to train Llama. 

• Christopher Farnsworth vs. Meta: The novelist filed a proposed class action alleging Meta used his and other authors’ books without permission for Llama training.

• Austin Beaulier vs. Meta: A 3D graphic artist alleges Meta used millions of models, including Creative Commons-protected assets, to train generative AI systems.

12. Caution, conclusion

There are no signs yet of a large-scale rebuild of Meta’s former media or content creator partnership teams, but nevertheless this writer is aware that Meta is holding frresh talks with publishers in several countries. 

For its part, Meta says it will “continue to add new partnerships and explore new features” to improve Meta AI. It probably has no choice but to deliver on that. For now, its approach to re-embrace media is selective and far from comprehensive. 

Media CEOs would be wise to reserve judgment on whether Meta is seriously back helping the news business until after it resolves Canada, re-signs Australia, and broadens its partnerships quest beyond its interesting but minimalist start. 

Above all, it is clear Meta isn’t returning to do a favour for the news ecosystem. 

It is returning because it needs high-quality, real-time content to improve its AI products. Its advertising business needs a top-shelf answer engine. To achieve that, its large language model needs journalism. 

And its return represents an embrace of copyright law that will have wider ramifications across the tech and media sectors. 

In this context, journalism has shifted from being a product category in its feeds to a data input for its AI effort. An overlap of Meta’s needs and publishers’ needs is where a true partnership will lie.

INMA will host an invitation-only executive briefing on AI deals, content value, copyright, and open standards at the World Congress of News Media on May 8 in Berlin under Chatham House rules. E-mail us if you’d like an invitation to the forum.

Banner photo: Adobe Stock By nikkimeel.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .