After 'ban' on Claude Code, Microsoft has more 'bad news' for Anthropic

Microsoft's Claude Code experience came to an abrupt end after the company issued a ban on the AI service in May, pushing engineers to transition to GitHub Copilot CLI. This move came as part of Microsoft's effort to rein in its AI spending before the fiscal year ended on June 30. The memo from Jay Parikh, executive vice president of Microsoft's CoreAI organisation, signaled that tokenmaxxing was not the primary goal, and urged teams to review their AI spending.

The memo also highlighted that OpenAI's GPT-5.6 would become the default model for internal use in place of Claude Code. The internal shift towards GitHub Copilot CLI sent a strong message to Anthropic, as the default-model line revealed that thousands of Microsoft engineers were inadvertently using Claude tokens without opting for the service, leading to unnecessary expenditures.

Parikh framed the transition as a move towards efficiency, emphasizing that the company aims to derive greater value from its token investment rather than simply reducing usage. Microsoft's AI token budget targets now put a ceiling on every division, with employees able to monitor their own spend via an internal dashboard. Although the relationship between Microsoft and Anthropic remains intact, with Claude models still accessible through Copilot CLI and Microsoft continuing to favor Anthropic within Microsoft 365 Copilot, the default shift to OpenAI models has undoubtedly hit Anthropic's enterprise coding workloads hard.