Google, Other Rivals Pull Out of Scale AI Over Security Concerns as Wang Joins Meta in Surprise Deal

Google breaks the links with the AI ​​scale after Meta’s shock announcement of the acquisition of a 49% stake in the AI ​​data reissue startup, reports Reuters, citing sources.

The agreement, which values ​​evolves at $ 29 billion, more than double its previous recovery of $ 14 billion – sent shock waves via Silicon Valley, triggering an exodus by large IA companies being distrusted to expose proprietary research and data sensitive to a direct competitor.

Familiar sources with development indicate that Google had planned to pay a scale of almost $ 200 million in 2025 for the training data marked by the crucial man for the development of its Gemini models. But this arrangement has now been rebuilt. The company, the largest client of the ladder, moves quickly to redirect contracts to rival suppliers. Already, Google has started talking to competitors in several scales this week in order to discharge most of its IA annotation needs.

This rapid change stems from the increase in concerns about META – Parent of Facebook, Instagram and Whatsapp – accessing the privileged data of the companies with which it is in competition directly in the arms race for artificial intelligence. Meta now has a powerful foot in the backbone of AI infrastructure, having almost half of the AI ​​on a scale and absorbing its CEO Alexandr Wang in its AI division.

The agreement was a silent coup

The meta-echelle partnership met quietly but quickly. Several sources suggest that the agreement has been orchestrated in recent months, because Meta has aggressively sought to strengthen its development of AI following disappointing journals of its Llama 4 model published in April. Although the model is promising, it did not manage to match the performance of the GPT-4 of Openai or Gemini from Google in key references, which encourages the fear that Meta was delaying.

To accelerate its progress, Meta has turned into AI of the scale, which has acquired since its foundation in 2016 as a high -end supplier of data sets marked by high quality humans – a crucial resource for the formation of advanced AI systems. Scale’s services are not cheap: some annotations of doctoral experts can cost more than $ 100 each. But its customers, including Google, Microsoft, Openai, Xai and the US government, were ready to pay precision.

With Wang, which should lead Meta AI efforts and several employees on a scale that also increases to the company, the agreement stimulates the alarm for rival companies. Many of them rely on the labeling scale not only the raw data, but also the outputs of the prototype model, internal prompts and rich examples in context which are at the heart of their development strategies. Now, these same companies fear that their crown jewels will be found in Meta’s line of view.

The backlash brings together the rhythm

The backlash was fast with Google, Microsoft, Elon Musk XAI and even Openai – which had already started to reduce its dependence on the scale ago – all moving away. Google in particular moves quickly to dismantle all key contracts with the scale. Although the exact calendar varies according to the agreement, the sources indicate that the quarter could be completed quickly due to the flexible structure of many data labeling transactions.

Labelbox, Turing, Handshake and Mercor – small competitors formerly overshadowed by the domination of Scale – now attend an increase in demand. The CEO of Labelbox plans to generate hundreds of millions of new revenues from defective customers. Handshake said his workload tripled within 24 hours of the meta-scaling. The CEO of Turing summarized the atmosphere in the industry: “Neutrality is no longer optional, it is essential.”

OPENAI, despite the expenses much less than Google on scale services, reiterated that it will continue to work with Scale, but stressed that the startup is only one of the many suppliers. Elon Musk’s XAI, meanwhile, is preparing to go out completely. Microsoft has not commented publicly, but it is also supposed to move its data labeling contracts.

A strategic bet for meta – and a risk of scale

The agreement is undoubtedly a victory for Meta. The appointment of Alexandr Wang should inject new technical vigor into the Meta AI roadmap. But for the AI ​​scale, the Meta Alliance could have a high cost. A large part of the company’s income – $ 870 million in 2024 – comes from providing services to companies that now consider it compromises. Unlike its government and automobile contracts, which can be isolated from competitive threats, the generative Lucrative AI sector which has propelled its growth is now on fragile terrain.

The company’s declaration following the agreement attempted to project stability, insisting that its “remains strong” business and that it is engaged in customer data protection. But he did not directly comment on the details of Google’s departure or the current customer exodus.

Beyond the immediate professional impacts, the meta-scale agreement should reshape the supply chain of the AI ​​industry. Companies have to recognize that control of data infrastructure – including labeling, annotation and fine adjustment processes – is just as critical as access to GPUs or large model architectures. This awareness pushes more laboratories to constitute internal data labeling teams and secure their own AI training pipelines, even at more cost.

Meta’s strategic bet on the acquisition of a direct line in this ecosystem is high risk and with a strong reward. Although it obtains a richness of internal capacities thanks to the scale and to Wang, the benefits can definitively alienate the meta meta meta at a time when the search for AI depends more and more on the confidence, interoperability and data security.

In the end, what Meta is gaining on a scale, he can lose in credibility, at least in the eyes of his fiercest IA rivals.