Microsoft and Meta are reportedly steering employees toward their own AI tools and reducing internal use of Anthropic’s Claude. The clearest explanation is organizational: both companies want tighter control over internal AI spending and are encouraging staff to use tools that fit their own workflows. The changes are significant, but they do not establish that Claude has become less capable—or that the companies have stopped offering Anthropic models to customers.
2
4
12
What is changing at each company
At Microsoft, projected annual spending on Anthropic technology for employees had reached at least $1 billion. That forecast was reportedly cut by more than a third, as staff were encouraged to try GitHub Copilot first.
4
24
At Meta, reported Claude Code use fell from roughly 60,000 employees to about 30,000. Spring layoffs contributed to the decline, but reports describe a broader shift toward Meta’s own coding tools. MetaCode had passed 30,000 users, while Muse Code had more than 6,000 internal users.
1
11
Why this is not a simple verdict on Claude
A company can limit employees’ use of an outside tool to manage costs or consolidate work around its own products without making a public judgment about the outside model’s quality. In these cases, the reported actions are budget and workflow changes: Microsoft is prioritizing Copilot, and Meta is promoting its own tools.
4
24
The distinction between internal use and customer offerings matters, too. The reported Microsoft cut concerns its own employees’ use of Anthropic technology; it does not mean Anthropic models have been removed from Microsoft’s enterprise platforms. Separate reporting said customer demand through enterprise platforms remained stable or growing.
7
12
The broader implication for AI providers
For an outside AI provider, fewer employee users at major companies could mean fewer internal seats or less direct usage. More broadly, the reports illustrate how access to enterprise customers may depend not only on a model’s appeal, but also on whether a company approves it, funds it, and makes it part of employees’ everyday tools. That is a potential commercial risk—not evidence that Anthropic’s overall revenue or growth is falling.
The reported pullback is therefore best read as a contest over control of workplace AI: who sets the default, pays for usage, and builds AI into employees’ routines. It is not, on its own, a measure of which model is best.