Google Pulls Back from Scale AI Amid Meta Stake & Data Leak Fears

 

Google's parent company, Alphabet, is reportedly set to significantly reduce its engagement with Scale AI, a pivotal AI model training data services startup.

 

This decision follows Meta's recent acquisition of a 49% stake in Scale AI, a deal that has seen the startup’s valuation surge to an estimated $29 billion, up from $14 billion prior to the investment.

 

The deal has sent ripples through the AI industry, notably unsettling Scale AI's existing clients.

 

Reuters, citing sources close to the matter, revealed that Google had earmarked approximately $200 million to spend with Scale AI in 2025 for human-labelled training data—a critical component for developing its Gemini AI model, a direct rival to ChatGPT.

 

In 2024, Google had already invested over $150 million with the firm.

 

Sources indicate that Google has already begun discussions this week with several of Scale AI's competitors.

 

This move aims to diversify its risk and shift workloads, particularly amidst anxieties that proprietary internal data from AI model developers could be inadvertently accessed by Meta through its significant shareholding and potential influence on Scale AI’s operations.
 

 

Scale AI's primary service involves deploying teams of human experts, including historians and PhDs, to prepare highly sensitive training data for advanced AI models.

 

This often includes confidential information like product prototypes, technical blueprints, or unreleased development strategies.

 

Consequently, numerous companies that compete directly with Meta in the burgeoning AI arena are now worried that continuing to work with Scale AI could inadvertently lead to their sensitive data falling into rival hands.

 

Google isn't the only tech giant re-evaluating its position. Microsoft and Elon Musk's xAI are also reportedly exploring ways to lessen their dependence on Scale AI.

 

OpenAI began a partial withdrawal several months ago, though its CFO recently confirmed Scale AI remains one of its data providers.

 

In an increasingly cut-throat AI landscape, technology companies are actively seeking "safer and more neutral" data solutions to prevent crucial information leaks.

 

Jonathan Siddharth, CEO of Turing – a rival AI data services firm – described the Meta-Scale AI deal as a "watershed moment."

 

He believes it has compelled tech companies to question whether they can truly trust third-party providers with direct ties to a competitor.
 

 

Historically, many firms chose data-labelling services from startups like Scale AI for their specialised expertise and speed.

 

However, with a direct AI market competitor like Meta now holding a substantial stake, Scale AI's perceived neutrality has become a critical vulnerability.

 

Siddharth emphasised that "neutrality" in this industry is no longer merely a matter of public image but a fundamental condition for client engagement and continued business relationships.

 

This shift in market dynamics presents significant opportunities for Scale AI's rivals, including Turing, Labelbox, Handshake, and Mercor, to secure new contracts, especially from AI companies demanding absolute business neutrality.

 

Handshake, for instance, reported a threefold surge in demand for its services from leading AI labs overnight.

 

Despite the potential upheaval from major client withdrawals, Scale AI maintains that its business structure remains robust, citing a strong client base from government and diverse industries, such as autonomous vehicle developers, who continue to use its services.

 

The company also reiterated its commitment to upholding the highest standards of client data protection, though it declined specific comment on Google's situation.

 

However, a core vulnerability for Scale AI is its reliance on a few key clients, notably Google, Microsoft, and OpenAI. Should these major clients reduce or shift their business elsewhere, Scale AI's revenue could experience a rapid decline.

 

This risk is further exacerbated by an emerging trend where large tech firms are moving away from outsourcing data management and are instead building in-house teams of data-labellers.

 

This shift aims to give them greater control over the security of sensitive data, mitigating the risk of confidential models or products being leaked. If this trend becomes mainstream, it could directly impact Scale AI's business model, which has long relied primarily on external service revenue.