Meta Restricts Employees From Using Claude and Codex to Prevent Model Distillation

According to Beating, a monitoring platform, Meta has recently restricted its AI engineering teams from using Anthropic's Claude Code and OpenAI's Codex, citing concerns over model distillation. The company fears employees may inadvertently incorporate external AI outputs into internal training data, potentially violating competitors' terms of service.

Meta now permits external AI tools only for routine tasks such as workflow setup, code organization, and test infrastructure building, with all outputs subject to strict human review. The company explicitly prohibits using external models to generate programming evaluations, identify code vulnerabilities, or design test scenarios. This effort supports Meta's push to consolidate operations around its proprietary coding assistant MetaCode.