Meta and Microsoft Limit Employee Use of Claude AI Tools

Meta and Microsoft have significantly curtailed employee use of Anthropic’s Claude AI as each shifts development toward internal coding assistants. Microsoft scaled back expected internal Anthropic spending - previously forecast above $1 billion annually - by more than a third and reduced monthly AI spending limits for many employees from $100,000 to roughly $10,000. Microsoft continues to offer Anthropic models to customers and client spending through its platforms is growing. At Meta, Claude Code users fell from about 60,000 to 30,000 as the company promotes MetaCode (over 30,000 internal users) and Muse Code (over 6,000 users), with Muse Code entering external tests in August. Despite fewer internal users, Meta still paid more than $105 million to Claude Code within a 28-day span, and Anthropic’s annualized revenue pace sits at about $65 billion - showing demand persists even amid retrenchment.

The shift reflects tightening budgets and intensifying competition rather than an outright break with Anthropic, but it does not eliminate security concerns. Previously discovered vulnerabilities in Claude Code and in Microsoft tooling (for example, a patched RoguePilot-style exploit) demonstrated risks like unauthorized command execution and API key theft; those flaws were fixed. These incidents reinforce the need for strict permissions, trusted project settings, and rigorous review of AI-generated actions as companies balance cost control, product strategy, and developer access.