Meta Platforms has drawn a hard line. Its Applied AI division now explicitly prohibits engineers from using Anthropic's Claude Code and OpenAI's Codex, two leading AI-powered coding tools.
The restriction, documented internally around June 29, 2026, aims to prevent a specific problem: rival AI models accidentally teaching Meta's own models their capabilities.
The Distillation Problem
The core concern is straightforward. When Meta engineers use these external tools, the outputs carry the proprietary DNA of Anthropic's and OpenAI's models. If that data then feeds into Meta's training pipelines, the company's models could absorb competitive capabilities they did not develop themselves.
This creates a potential breach of service agreements with both Anthropic and OpenAI, which typically prohibit using their outputs to train competing models.
Teams have been instructed to halt tasks involving these tools, enforce human oversight on remaining workflows, and scrub any prior external outputs from training data and benchmarks.
The Scale Was Enormous
Before the ban, internal tracking showed staggering usage. An internal system, nicknamed the "Claudeonomics" dashboard, recorded 60 trillion tokens consumed within a single 30-day period.
Microsoft reportedly also canceled a majority of its Claude Code licenses by June 30, 2026, citing excessive token consumption as a primary driver.