A major concern is that Scale, a key provider of data labeling for big tech firms and AI startups to train their models, might share insights into the types of data used by top AI companies to develop their most advanced technologies with Meta.

As one former employee of Scale told Forbes: “They all want to cut ties with Scale now. The entire business model of Scale could collapse once it becomes part of Meta.”

OpenAI, one of Scale’s largest clients, has already started reducing its collaboration with the company, according to four sources familiar with its operations; two said this process has been ongoing for months and OpenAI has been evaluating alternative partners. OpenAI declined to comment.

This unexpected move, which would value the company at $28 billion, left internal teams at Scale in disarray and confused, as noted by the former employee. Some employees are worried about what level of visibility Meta might gain regarding previous projects, even though most agreements require data deletion after project completion, he added.

Scale declined to comment.

Smaller competitors of Scale have already begun positioning themselves and are eager to attract customers concerned about potential conflicts of interest involving data and privacy. Invisible Technologies co-founder Francis Pedraza told Forbes his firm is committed to remaining independent. Turing, a startup that already provides training data to OpenAI, Anthropic, and Google, sees this deal as an opportunity to be “Switzerland” and serve as a neutral supplier of data to leading AI labs. CEO Jonathan Sidharth told Forbes, “Clients prefer working with someone who is impartial and can support all labs equally.” One investor backing a competitor of Scale mentioned the deal could create new opportunities for other companies to “fill the gap left by Scale AI.”

Access to high-quality human-labeled data is essential for training powerful AI models, so much so that data has become a competitive advantage for AI giants like OpenAI and Anthropic in the race for dominance. In 2024, Scale AI generated $870 million in revenue and established itself early in the data labeling market by supplying vast amounts of labeled data to companies such as Cohere, OpenAI, and Microsoft. However, if Meta ends up owning nearly half of Scale, that dynamic will change.

Scale built its business using a large workforce of clickworkers, primarily based overseas, who provided context for massive datasets to help train AI models. But this kind of work has become increasingly standardized. As Kevin Guo, co-founder of a startup that exited the data labeling space, said in a 2023 interview: “Anyone who assembles a team can compete with you, and it quickly comes down to price.”

One senior executive in the AI industry described Scale as “the bulk food section of the AI training market.” Multiple sources familiar with Scale’s operations told Forbes the company has struggled with quality issues. “They tend to overpromise and oversell, and often underdeliver,” said the former employee.

Wang, currently one of the youngest self-made billionaires at age 28 with a net worth estimated at $3.6 billion, will lead a new lab at Meta focused on developing so-called superintelligence—an AI system capable of surpassing human capabilities, according to The Information. To build this team, Zuckerberg is reportedly offering researchers from Scale AI, OpenAI, Anthropic, and Google DeepMind more than $10 million annually to join the new lab. Bloomberg reported that Zuckerberg is personally involved in assembling the team, going as far as setting up a WhatsApp group named “Recruiting Party,” reaching out directly to prospects, and rearranging office layouts so researchers sit close to him.

The deal has yet to finalize and may face regulatory hurdles. If completed, it would result in significant financial gains for Wang and early investors in Scale AI, including Accel and Index Ventures. However, the future of Scale remains uncertain. “This is great for Alex and early investors but terrible for everyone else, including current and former employees,” said a former senior employee at Scale AI. “It’s unclear how this deal benefits Scale.”

In recent years, the world’s biggest tech companies have been fiercely competing to achieve AI dominance. Meta, which launched its AI lab back in 2013, has struggled to keep pace with rivals like Google, OpenAI, and Anthropic despite its open-source Llama models. Its reputation took a hit in April when it was accused of artificially inflating benchmark scores for its Llama 4 models (a claim Meta denied). Bringing on someone as prominent as Wang could provide a boost to the company’s AI efforts.

Zuckerberg has also pushed Meta toward defense contracts, and Scale’s government division could align well with those goals (though, as Forbes previously reported, this segment has had difficulty gaining momentum).

Meta joins a growing list of companies poaching CEOs from high-profile AI startups. Last year, Microsoft hired Mustafa Suleyman, a founder of DeepMind, along with his senior team from Inflection, the AI lab he launched in 2022, effectively leaving the company hollowed out. Months later, Amazon signed a deal with Adept to license its technology and hire CEO David Luan and his founding team. Similarly, Google re-hired researcher Noam Shazeer—who co-invented the transformer architecture that powers generative AI during his time at Google—from his startup Character.AI.

Matt Fitzgerald, CEO of Invisible Technologies, noted that as models grow more complex, they require higher levels of expertise. Companies like Scale, Turing, and Invisible have shifted away from low-paid clickwork to focus on more advanced tasks handled by PhDs and highly educated professionals. He said Meta’s move reinforces the importance of human input in AI training. “This is a long-term bet that human involvement will remain critical for many years to come,” Fitzgerald said.

John Paczkowski, David Jeans and Iain Martin contributed reporting.

More from Forbes

ForbesHow Alexandr Wang Turned An Army Of Clickworkers Into A $7.3 Billion AI UnicornBy Kenrick CaiForbesMeta Is Recruiting Former Pentagon Officials As It Ramps Up Military AmbitionsBy David JeansForbesDefense Contractor Scale AI Quietly Scrapped Deal With Chinese-Owned TikTok Over Security ConcernsBy David JeansForbesA Growing Side Hustle For American College Grads: Fixing AI’s Wrong AnswersBy Richard Nieva

The above is the detailed content of Scale AI's Business Could Collapse If Meta Buys A Stake And Hires Its CEO. For more information, please follow other related articles on the PHP Chinese website!