Microsoft and Meta Cut Internal Claude Use for In-House Tools
Microsoft and Meta are sharply cutting employees' internal Claude use and steering staff to their own coding tools, while customer use via Microsoft still grows.
Microsoft and Meta are sharply reducing employees' internal use of Anthropic's Claude and steering staff toward their own coding tools, according to reports citing The Information. Microsoft's expected internal Anthropic spend was cut by more than a third from a prior projection that sources describe as above $1 billion a year or, in one account, at least $1 billion; executives told employees to use GitHub Copilot and OpenAI models instead. Sources differ on the scope of new caps: one account says the cloud division's monthly per-employee budget fell from $100,000 to about $10,000, while others say many cloud and AI staff, in most cases, saw monthly limits drop by those amounts. Meta's Claude Code users fell from roughly 60,000 to about 30,000 as Meta promoted MetaCode and Muse Code, even after spending more than $105 million on Claude Code in one 28-day period; cost control, limits on Anthropic's access to Meta training data, and rivalry with tools such as Claude Cowork are also cited. Customer access to Claude through Microsoft products is not being withdrawn, and customer use and spending via those products are reportedly still growing. One account additionally cites previously patched Claude Code flaws CVE-2025-59536 and CVE-2026-21852, which it says enabled unauthorized execution and API key theft, plus Microsoft's RoguePilot Codespaces flaw, as reasons to govern agent permissions.
- Microsoft's projected internal Anthropic spend was cut by more than a third from a prior figure sources put above $1 billion a year, or at least $1 billion in one account.
- Monthly per-employee caps fell from about $100,000 to about $10,000; one account limits that to the cloud division, while others say many cloud and AI staff, in most cases, were affected.
- Microsoft executives told employees to use GitHub Copilot and OpenAI models instead of Claude.
- Meta's Claude Code users fell from roughly 60,000 to about 30,000 as Meta promoted MetaCode and Muse Code.
- Meta spent more than $105 million on Claude Code in one 28-day period.
- Customer access to Claude through Microsoft products is not being withdrawn, and customer use and spending via those products are reportedly still growing.
- Cost control, limits on Anthropic's access to Meta training data, and rivalry with tools such as Claude Cowork are also cited.
- One account cites previously patched Claude Code flaws CVE-2025-59536 and CVE-2026-21852, which it says enabled unauthorized execution and API key theft, plus Microsoft's RoguePilot Codespaces flaw, as reasons to govern agent permissions.
Coverage timelineoldest first · each row is one article
- · 3d agoMeta and Microsoft pull back from Claude as Anthropic transforms from partner into competitor
The Decoder· 66
Microsoft and Meta are sharply cutting internal Claude use as Anthropic becomes a competitor.
- · 2d agoMeta and Microsoft are Actively Cutting Employee Use of Claude AI
Cyber Security News· 52
Meta and Microsoft are cutting employees' Claude use while steering staff to their own coding tools.
- · 2d agoMeta and Microsoft Reduce Internal Use of Claude AI
GBHackers· 48
Meta and Microsoft are cutting internal Claude use while pushing their own coding assistants.
Vulnerabilities in this storyAll →
- CVE-2025-595368.726%Code Injection in Anthropic Claude Code Startup Trust Dialog (pre-1.0.111)published · Anthropic Claude Code
- CVE-2026-218525.325%API Key Exfiltration via Attacker-Controlled Base URL in Anthropic Claude Code