Meta Clamps Down on Engineers Using Claude Code and Codex
Meta restricts its AI division from using Anthropic and OpenAI coding tools over fears of accidental model distillation.
Meta is locking down access to rival AI coding tools. Internal documents reveal the company is imposing strict limits on how engineers in its applied AI division can use Anthropic's Claude Code and OpenAI's Codex.
The core fear: inadvertent distillation. That's when a company's own models accidentally absorb capabilities from competitor AI systems through routine usage — essentially letting rivals' intelligence leak into your own stack.
The restrictions come as Meta tries to reduce its dependency on expensive external AI coding tools. It's a tricky balancing act. Engineers need powerful coding assistants to stay productive, but leaning on competitors' products creates both cost problems and intellectual property headaches.
Meta wants to build its own AI ecosystem from the ground up. Letting Anthropic and OpenAI's fingerprints seep into that foundation defeats the purpose entirely.