During the sweltering months of summer, Meta made headlines by investing an extraordinary $14 billion into the data labeling powerhouse Scale AI while simultaneously recruiting its prodigious 28-year-old founder, Alexandr Wang. In the wake of this transformative deal, major technology partners—among them OpenAI and Google—abruptly paused their collaborations with the startup. Within this climate of uncertainty, an anxious Scale AI contractor turned to ChatGPT, a system with which he had become intimately familiar through professional testing for security weaknesses, to seek an unvarnished prediction about the company’s prospects. The chatbot’s analytical response was far from reassuring. It warned that within two years, Scale AI might cease to exist as an independent and credible organization, forecasting that its technological infrastructure would be absorbed into Meta’s internal ecosystem, its established client network would disperse, and its once-valued role as a neutral evaluator or external red-teaming service would effectively come to an end.
This ominous dialogue did not remain private for long. The contractor shared the conversation with colleagues inside Scale AI, sparking a wave of unease. According to internal exchanges reviewed by Business Insider, one worker confessed in resignation that they were already seeking other opportunities, describing the company as a “ticking time bomb” poised to implode. ChatGPT’s bleak outlook echoed a media narrative that had been unfolding for months: the meteoric rise—and now potential decline—of a company once hailed as the unsung engine powering the artificial intelligence revolution. Only a short time ago, Scale AI had been the destination for every major tech organization vying for AI dominance, serving as the preferred partner to test, refine, and fortify large language models against weaknesses and bias. Yet as the months progressed, its once lofty valuation and glittering reputation had begun to dim, weighed down by investor skepticism, internal dissatisfaction, and the encroachment of nimble competitors.
The backbone of Scale AI’s success—its global army of human data labelers—has become restless and disillusioned. Many contractors have voiced frustration over significant reductions in compensation, drawn-out unpaid onboarding phases for new projects, and an overall scarcity of available work. These grievances have fueled a growing exodus from the platform, according to both confidential communications and multiple interviews conducted with current and former contributors. The decline in morale is visible even in Scale’s internal communication channels. Activity on the central chat forum for Outlier, the company’s massive gig-based platform boasting more than one hundred thousand registered taskers, has markedly decreased since Meta’s investment. Screenshots reviewed by Business Insider depict a once-vibrant message board that previously amassed hundreds of comments now garnering only a few scattered replies each week.
Several workers have described their experiences in stark terms. One contributor reported dedicating nearly forty hours in a single month to unpaid onboarding activities without securing any actual compensated assignments, quickly discovering that rival platforms, such as the competing startup Mercor, pay contractors for similar preparatory work. Another, Elizabeth Boyd, observed a dramatic erosion of her earning power—where she had once made around fifty dollars per hour, her effective rate had plummeted to roughly twenty. In one extreme case, a project advertised at twenty dollars per hour allowed for only three minutes of labor every forty-eight hours, generating payouts barely reaching a single dollar.
Despite these disquieting accounts, Scale AI’s leadership has maintained a confident public stance. Company spokesperson Joe Osborne asserted that internal financial metrics demonstrate robust performance and a trajectory of renewed strength. In an email statement, Osborne claimed the current quarter was on track to become the most lucrative of 2025. He noted that the firm’s data services division was now more profitable than prior to the Meta partnership, while its applications arm—serving major Fortune 500 enterprises and government entities—had doubled its revenue between the year’s first and second halves. He further emphasized that user activity on the Outlier platform had increased since the investment, adding that payment levels are calibrated to the skill set required for each distinct project and that contributors are always informed of compensation rates before accepting an assignment.
Scale AI, perhaps in an effort to regain its edge, has broadened its strategic ambitions. The company has plunged into emerging areas such as robotics, announcing the creation of a dedicated laboratory intended to satisfy the rapidly expanding demand for training data used in robotic systems. Simultaneously, it has intensified its focus on contracts with the U.S. military and various governmental organizations, securing agreements worth up to $199 million since entering its partnership with Meta.
Investor sentiment surrounding the company remains starkly divided. Optimists argue that despite Meta’s substantial shareholding, Scale AI continues to function largely autonomously. One current investor described Meta’s approach as mostly hands-off, adding that with roughly a billion dollars still present on its balance sheet, Scale is not seeking external fundraising and retains the possibility of an eventual initial public offering. However, others interpret the situation less favorably. Some observers liken the company’s condition to that of a gutted fish, pointing out that the $29 billion valuation implied in Meta’s deal has been sharply devalued in private markets. Noel Moldvai, CEO of Augment, told Business Insider that his trading platform had once facilitated multimillion-dollar transactions in Scale AI equity, only for that market activity to evaporate after the Meta transaction as investors waited for evidence of recovery. Although trading volumes have recently rebounded somewhat, valuations have sunk to between $9 billion and $15 billion. Moldvai suggested that Meta’s real motivation may have been to recruit Alexandr Wang himself rather than to acquire Scale’s assets outright—though he added that a future rebound is not impossible. On another secondary marketplace, Caplight, share pricing suggests an even more modest valuation of roughly $7.3 billion, a figure Osborne disputed as inaccurate since there have been no verified share sales at that price and comparable firms would command higher multiples.
Should Scale AI fail to reverse its current trajectory, analysts warn that it could join the ranks of once‑promising startups that, following an influx of capital from industry behemoths, became listless entities—“zombie” companies, surviving but devoid of their former vigor. In more optimistic public communications, however, Scale AI initially portrayed Meta’s infusion as a transformative partnership—a mutually beneficial alliance promising fresh resources, long-term security, and an expanded horizon of opportunities. That narrative was also echoed inside the company. Yet optimism swiftly gave way to restructuring: only weeks after the deal closed, Scale announced layoffs impacting fourteen percent of its 1,400-person full-time workforce. Osborne explained that these cuts were designed to restore profitability to the company’s data division, which, according to him, has since achieved that goal. The reductions did not stop there. In September, a dozen contractors from Scale’s red team—responsible for testing and probing AI systems for weaknesses—were terminated on performance grounds. Former team members, however, contended that dwindling project volume was the real reason behind their dismissal. Later that same month, Scale shuttered a Dallas-based group of contractors engaged in general AI-related assignments as the company pivoted toward more specialized work, a move Osborne framed as reflective of a broader industry trend rather than an isolated retrenchment.
Meanwhile, competition within the AI training sector has intensified dramatically. Numerous emerging startups have seized upon Scale AI’s sudden vulnerability, luring away both personnel and clients. New entrants like Surge AI and Mercor have been raising vast sums at startlingly high valuations—Surge reportedly reaching $24 billion, while Mercor, helmed by its youthful founders in their early twenties, secured a $350 million round at a valuation near $10 billion. Notably, Mercor has already captured at least one significant training project for Meta, which holds a 49% stake in Scale. In response, Scale filed a lawsuit in California, accusing Mercor of poaching its customers by recruiting a former sales employee, allegations Mercor denies. These tensions underscore the broader turbulence defining Scale’s current environment—an ecosystem where loyalty is fleeting, competition ruthless, and reputation fragile.
Some investors have expressed frustration that Scale’s leadership allowed competitors, particularly Surge AI, to overtake it in revenue generation despite Surge’s having raised no external funding. Brendan Foody, CEO of Mercor, emerged as one of Scale’s most outspoken critics, publicly accusing the company of compromised quality standards and declining pay scales. In his account, Scale’s breakneck expansion eroded its commitment to excellence, leading to what he called a “loss of focus on product and on scaling quality.” Scale AI, for its part, rebutted those accusations, asserting that its quality metrics have reached record highs. Yet similar criticisms have circulated among former insiders. Tammy Hartline, previously a project consultant at Scale, remarked that as the company grew exponentially, its processes increasingly prioritized volume over precision. “Spam and low-quality data became accepted as a routine cost of doing business,” she explained, observing that she ultimately joined Mercor as a result.
Beyond internal and competitive challenges, Scale AI has also grappled with a series of security mishaps predating the Meta deal. Investigations revealed that the company habitually employed publicly accessible Google Docs to coordinate assignments for top-tier clients—including Meta itself, Google, and xAI—leaving confidential AI training data exposed to anyone possessing the shared links. These files contained not only sensitive corporate information but also personal details such as contractor payment data and email addresses. Osborne acknowledged the issue, asserting that the company conducted an exhaustive internal probe and subsequently disabled users’ ability to publicly share any company-managed documents. While other AI service providers like Surge AI have also suffered similar oversights, Business Insider’s review of the exposed documents demonstrated that many ongoing projects—particularly one commissioned by Google—faced persistent quality and security shortcomings in both 2023 and 2024. Thousands of individual contributors were flagged as suspected spammers or cheaters, with spreadsheets internally titled “Good and Bad Folks” and “Suspicious Non‑US Taskers.” Following related reporting, Meta took action to dismantle more than forty online groups engaged in buying and selling Scale AI training accounts. Osborne nevertheless maintained that Scale’s data‑quality benchmarks have never been higher.
Despite the string of difficulties, the company has recently begun to resolve certain legal entanglements that once weighed heavily upon it. It has agreed to settle multiple lawsuits filed by former California-based contractors who had alleged wage underpayment and misclassification of their employment status, a step aimed at closing a contentious chapter in its labor relations. Scale no longer engages gig workers from the state. The overarching question that looms over all these developments is whether Scale AI—once emblematic of Silicon Valley’s frontier of machine learning—can reclaim vitality within the competitive ecosystem it helped to shape. For many former employees watching from afar, the answer may come too late to alter their personal fortunes, but the outcome will nonetheless offer a potent case study in how swiftly dominance in the modern AI landscape can erode when ambition collides with corporate upheaval.
Sourse: https://www.businessinsider.com/pay-cuts-poaching-pivoting-inside-scale-ai-meta-2025-12
