Meta, the parent company of Facebook, Instagram, WhatsApp and Messenger, is facing antitrust proceedings that could limit its ability to develop AI amid a field of competitors.

First filed in 2021, the Federal Trade Commission (FTC) alleges that Meta’s strategy of absorbing firms — rather than competing with them — violates antitrust laws. If the court rules against Meta, it could be forced to spin out its various messenger services and social media sites into independent companies.

The loss of its stable of social media companies could harm Facebook’s competitiveness not only in the social media industry but also in its ability to train and develop its proprietary Llama AI models with data from those sites.

The trial could take anywhere from a couple of months to a year, but the outcome will have lasting consequences on Meta’s standing in the AI race.

Meta’s antitrust case and its effect on AI

The FTC first opened its complaint against Meta in 2020 when the firm was still operating as Facebook. The agency’s amended complaint a year later alleges that Meta (then Facebook) used an illegal “buy-or-bury” scheme on more creative competitors after its “failed attempts to develop innovative mobile features for its network.” This resulted in a monopoly of the “friends and family” social media market.

Meta founder and CEO Mark Zuckerberg had the chance to address these allegations on April 14, the first day of the official FTC v. Meta trial. He testified that only 20% of user content on Facebook and some 10% on Instagram was generated by users’ friends. The nature of social media has changed, Zuckerberg claimed.

“People just kept on engaging with more and more stuff that wasn’t what their friends were doing,” he said — meaning that the nature of Meta’s social media holdings was sufficiently diverse.

The FTC alleges that Meta identified potential threat competitors and bought them up. Source: FTC

At the time of the FTC’s initial complaint, Meta called the allegations “revisionist history,” a claim it repeated on April 13 when it stated the agency was “ignoring reality.” The company has argued that the purchases of Instagram and WhatsApp have benefited users and that competition has appeared in the form of YouTube and TikTok. 

If the District of Columbia Circuit Court rules against Meta, the global social media giant will be forced to unwind these services into independent firms. Jasmine Enberg, vice president and principal analyst at eMarketer, told the Los Angeles Times that such a ruling could cost Meta its competitive edge in the social media market.

“Instagram really is its biggest growth driver, in the sense that it has been picking up the slack for Facebook for a long time, especially on the user front when it comes to young people,” said Enberg. “Facebook hasn’t been where the cool college kids hang out for a long time.”

Such a ruling would also affect the pool of data from which Meta can draw to train its AI models. In July 2024, Meta halted the rollout of AI models in the European Union, citing “regulatory uncertainty.” 

The pause came after privacy advocacy group None of Your Business filed complaints in 11 European countries against Meta’s use of public data from its platforms to train its AI models. The Irish Data Protection Commission subsequently ordered a pause on the practice until it could conduct a review. 

Related: Meta’s Llama 4 puts US back in lead to ‘win the AI race’ — David Sacks

On April 14, Meta got the go-ahead to use public data — i.e., posts and comments from adult users across all of its platforms — to train the model. If these firms dissolved into separate companies, with their own organizational structures and data protection policies and practices, Meta would be cut off from an ocean of data and human communication with which its AI could be improved. 

Andrew Rossow, a cyberspace attorney with Minc Law and CEO of AR Media Consulting, told Cointelegraph that in such an event, “companies would most likely control their own user data, and Meta would be restricted from using it unless new data-sharing agreements were negotiated, which would be subject to regulatory scrutiny and user/consumer privacy laws.”

However, Rossow noted that it wouldn’t be a total loss for Meta. Zuckerberg’s firm would retain the wealth of data from Facebook and Messenger. It could continue to use “opt-in” data from consumers who allow their posts to be used for AI training, and it could also employ synthetic data sets as well as third-party and open data.

Meta, the AI race and data protections

The race to unseat OpenAI and its ChatGPT model from AI dominance has grown more competitive in the last year as DeepSeek joined the fray and Meta launched the fourth iteration of its open-source Llama model. 

In addition to training new models, major AI development firms are investing billions in new data centers to accommodate new iterations. In January 2025, Meta announced the construction of a 2-gigawatt data center with more than 1.3 million Nvidia AI graphics processing units. 

Zuckerberg wrote in a post on Threads, “This will be a defining year for AI. In 2025, I expect Meta AI will be the leading assistant serving more than 1 billion people […] To power this, Meta is building a 2GW+ datacenter that is so large it would cover a significant part of Manhattan.”

Illustration of the data map coverage. Source: Mark Zuckerberg

His announcement followed the $500-billion Stargate project, which would see massive investment in AI development led by OpenAI and SoftBank, with Microsoft and Oracle as equity partners. 

Related: Trump announces $500B AI infrastructure venture ‘Stargate’

Amid this competition, AI firms are looking for broader and more varied sources of data to train their AI models — and have turned to dubious practices in order to get the data they need. In order to stay competitive with OpenAI when developing its Llama 3 model, Meta harvested thousands of pirated books from the site LibGen. According to court documents in a case pending against Meta, Llama developers harvested data from pirated books because licensing them from sources like Scribd seemed “unreasonably expensive.” 

Time was another perceived motivator for using pirated works. “They take like 4+ weeks to deliver data,” one engineer wrote about services through which they could purchase book licenses.

The practice is not limited to Meta. OpenAI has also been accused of mining data from pirated work hosted on LibGen. 

Rossow suggested that, “to ensure lasting impact — beyond short-term profit,” Meta would do well to “prioritize investment in advanced data collection, rigorous auditing and the implementation of privacy-preserving and encryption-based technologies.”

By focusing on transparency and responsible practices, “Meta can continue to genuinely advance AI capabilities, rebuild and nurture long-term user trust, and adapt to evolving legal and ethical standards, regardless of changes to its platform portfolio.”

What a ruling for the FTC would mean

Litigation is now hitting tech firms from all sides as they face allegations of privacy violations, copyright law infringement and stifling competition. Major cases like those facing Google, Amazon and Meta that have yet to play out will decide how and whether these firms can proceed as they have, defining the guardrails for AI development as well. 

Rossow said that the current antitrust case against Meta could decide how courts interpret antitrust law for tech firms, spanning tech mergers, data usage and market competition. It would also signal that courts are “willing to break up tech conglomerates” when issues of smothering competition are involved, while at the same time, “taking current precedent a step further in harmonizing it with the laws of cyberspace.”

Magazine: Memecoin degeneracy is funding groundbreaking anti-aging research