Meta contractors posed as teens to test rival chatbots, Wired reports

Hundreds of contractors working on a project for Meta ↘ 0.04% were instructed to pose as children and probe competitor chatbots — including Google's Gemini Alphabet Inc. ↘ 1.02% and OpenAI's ChatGPT — with prompts involving suicide, sex, and drugs, according to a report published Sunday by WIRED.x+1

The contractors, based in Kenya, sent images including pills, knives, nooses, and a medical diagram of a gynecological procedure as part of the testing, the report found. The prompts were designed to push rival AI systems to their limits and expose safety failures in how those chatbots respond to minors seeking dangerous content.tiktok+3

A Pattern of Outsourced Testing

The revelation adds a new dimension to ongoing scrutiny of how tech companies test and safeguard their AI products, particularly around children. Meta has faced repeated criticism over its own chatbots' interactions with minors, including an internal red-team assessment that showed a 66.8% failure rate in blocking child sexual exploitation content and a 54.8% failure rate on suicide and self-harm prompts. The company paused teen access to AI companion characters in January 2026 amid legal pressure.axios+1

A Swedish news outlet previously reported that Meta contractors in Kenya had been involved in the testing, which targeted how competitors handled sensitive conversations with users claiming to be underage.instagram

AI Replaces Human Moderators

The report comes as Meta accelerates its pivot away from human content moderation. According to the Financial Times, reported last week, the company plans to replace more than 90% of its content review workforce with large language models by the end of 2026. Meta has already shifted roughly half of all human review requests to AI this year, claiming the systems make 13% fewer mistakes than human reviewers while catching 10% more policy violations.socialmediatoday+2

The human cost of this transition has already materialized. In April, Nairobi-based outsourcing firm Sama issued redundancy notices to 1,108 employees after Meta formally terminated a major engagement. That contract termination followed whistleblower reports from Kenyan workers who described disturbing content they were asked to review through Meta's smart glasses footage.washingtonpost+2

Meta's dual approach — using low-paid contractors abroad to stress-test competitors while simultaneously eliminating moderation jobs in favor of automation — underscores the tensions at the heart of the AI safety debate as companies race to deploy increasingly powerful systems with fewer human guardrails.