Woofun AI reports that Meta has imposed strict limitations on its AI engineering team regarding the use of Anthropic's Claude Code and OpenAI's Codex. The company requested a pause on certain related tasks to enhance compliance reviews, driven by concerns that employees are overly relying on external tools to develop internal alternatives. This reliance risks competitor model outputs seeping into Meta's own training data, a practice known as 'model distillation' that may violate competitors' terms of service and trigger conflicts with partners.

Currently, Meta permits employees to use external AI only for setting up workflows, organizing code, and building testing infrastructure as part of routine work, with all outputs requiring thorough manual review. The use of external model-generated programming challenges to assess internal models is strictly prohibited, as is leveraging external AI to discover code vulnerabilities or brainstorm testing tasks. This move supports Meta's efforts to reshape its internal toolchain and control AI operations costs, which reach tens of billions of dollars, by promoting its in-house coding assistant MetaCode.