[Meta Restricts Employees from Using Claude and Codex to Prevent 'Model Distillation' Violations] According to monitoring by Beating, Meta is strictly limiting its AI engineering teams from using Anthropic's Claude Code and OpenAI's Codex. The company has even requested the suspension of certain related tasks to strengthen compliance reviews. Meta is concerned that employees might overly rely on external tools to develop internal alternatives, leading to the output of competing models being incorporated into its own training data. This practice of 'model distillation' could potentially violate the terms of service of competitors and cause serious conflicts with partners. Currently, Meta only allows employees to use external AI for routine tasks such as setting up workflows, organizing code, and building testing infrastructure, with all outputs requiring strict manual review. However, it strictly prohibits the use of external models to generate programming challenges for evaluating its internal models, as well as using external AI to identify code vulnerabilities or design testing tasks. This move is also part of Meta's effort to reshape its internal toolchain and control its internal AI operational costs, which amount to billions of dollars, by promoting its in-house coding assistant MetaCode (formerly known as DevMate). [Original Link]