Meta has placed strict limits on how engineers in its Applied AI division can use Claude Code and Codex, amid concerns that extensive use of AI coding assistants could inadvertently lead to the distillation of proprietary capabilities from Anthropic’s and OpenAI’s models. The growing adoption of AI coding tools has turned model distillation from a theoretical risk into a formal enterprise policy concern at one of the world’s largest AI labs by headcount. This marks one of the first documented instances of a major frontier AI lab restricting employee access to rival AI coding assistants on the grounds of preventing model distillation. Model distillation, also known as knowledge distillation, is an AI training technique in which a smaller, more efficient model learns to replicate the behaviour of a larger, more powerful model. Using a teacher-student framework, the smaller “student” model is trained on the outputs generated by the larger “teacher” model, allowing it to achieve similar performance while requiring significantly fewer computing resources.

Full report : Meta is placing strict limits on how engineers in its applied AI division can use Claude Code and Codex, fearing inadvertent distillation.