Meta has quietly banned its engineers from using two of the most popular AI coding tools on the market: Anthropic's Claude Code and OpenAI's Codex.
The policy, effective as of June 2026, is designed to protect the company's proprietary AI training data and internal codebases from potential exposure.
The core issue is how these tools function. They operate by sending code context to external servers for processing. This means when a Meta engineer uses Claude Code to debug a script, portions of the company's codebase are transmitted outside Meta's secure environment.
This move follows updates to Anthropic's consumer terms in late 2025, which now allow for opt-in training on certain datasets. That change likely heightened security and legal scrutiny at major tech companies.
Both Anthropic and OpenAI have been competing aggressively, with their pro-tier subscriptions priced around $20 per month. This accessibility has driven rapid adoption in engineering teams, amplifying concerns about data leakage at scale.
The restrictions highlight a critical challenge for the AI coding tools market. Developers like Anthropic and OpenAI must create enterprise-grade solutions that meet strict data residency and privacy requirements. While both offer enterprise products, Meta's internal ban suggests these may not yet address all security concerns.
Meta has not issued a public comment on the specific policy. The decision is being handled as a routine internal security matter.