For years, the work of deciding what billions of people could see on Facebook, Instagram, WhatsApp, and Threads was carried out by human contractors, reviewing flagged posts across dozens of languages and making judgment calls that shaped public discourse. That model is now being fundamentally restructured.
Meta Platforms is working to replace human content moderators with large language models as part of a broader effort to reduce operating costs, while Chief Executive Mark Zuckerberg directs spending towards AI development. The pace of the transition is notable. Meta has already replaced about 50% of human review requests with large language models this year. According to reporting by the Financial Times, that figure could climb to over 90% for certain content types before the end of 2026.
The financial rationale is straightforward. Meta spent roughly $5 billion on content moderation in 2025, with the bulk going to third-party contractors who review millions of posts daily. Automating that function would free up significant capital, which Zuckerberg has indicated will be redirected towards AI infrastructure and what he describes as the development of “personal superintelligence.”
The Workforce Behind the Platforms Faces Disruption
The transition is already affecting the large outsourcing ecosystem that Meta has long relied upon. The company’s shift represents a fundamental change from its dependence on third-party vendors like Accenture and Cognizant, which currently employ an estimated 15,000 to 20,000 contract moderators reviewing content flagged by Meta’s algorithms.
The downstream effects are becoming visible across the industry. Scale AI cut about 14 percent of its staff and 500 global contractors in 2025, and major outsourcers like Cognizant and Accenture, which together employ more than 15,000 workers on Meta contracts, have announced workforce reviews.
Meta has clarified that it will still rely on experts to design, train, and oversee its AI content enforcement systems, and that humans will remain involved in the most complex, high-impact decisions involving law enforcement and appeals. Routine enforcement, however, is increasingly becoming the domain of automated systems.
For India, which accounts for over 50 crore users across Meta’s family of apps and is the company’s single largest market by user base, the stakes of this transition are particularly significant. Content moderation in India has historically been undermined by the limited capacity of Meta’s algorithms and human moderators to recognise many of India’s 22 officially recognised languages. Delegating that responsibility further to AI systems raises practical questions about accuracy, cultural context, and linguistic diversity.
Accuracy Concerns and Governance Gaps
The operational risks of large-scale AI moderation have already surfaced in documented cases. Meta has laid off significant portions of its human content moderation teams and replaced them with AI enforcement systems that flag, restrict, and permanently disable accounts automatically, with no human review involved in the vast majority of cases.
Thousands of users have reported having their accounts disabled after AI incorrectly flagged ordinary content, with no viable path to contest the decision and no human moderator reviewing it before enforcement. The problem has drawn scrutiny even from within Meta’s own governance structures. In June 2026, the company’s Oversight Board confirmed that account bans lack due process and transparency, noting that many appeared to have been made entirely by automated systems, with users receiving no meaningful explanation.
On the product development side, the automation is equally far-reaching. Up to 90% of all risk assessments for new product features will soon be automated, meaning critical updates to Meta’s algorithms and changes to how content is shared across its platforms will largely be approved by AI systems, rather than human evaluators tasked with examining potential unintended consequences.
Regulatory Scrutiny and the Road Ahead
The shift arrives as Meta continues to scale back other moderation commitments. The company’s decision to dismantle its third-party fact-checking programme in the United States earlier this year added to concerns already being voiced by regulators and civil society organisations. In April 2025, Meta’s own Oversight Board publicly questioned whether the company’s rollback of content moderation policies had gone too far in dismantling safeguards for users.
Meta has stated its commitment to publishing ongoing enforcement data, enabling users, regulators, and researchers to assess the company’s performance over time. How rigorously that data is scrutinised, and whether it prompts meaningful regulatory action, will likely determine how this transition unfolds for the hundreds of millions of users across India and beyond who rely on these platforms daily.
