Anthropic merges Claude Chat and Cowork for Enterprise, routing heavy tasks into isolated cloud sandboxes

Anthropic's latest Enterprise change removes the line between talking to Claude and having Claude do work. Until now, Chat and Cowork were separate surfaces. Under the unified experience, a user begins in an ordinary conversation, and Claude moves the session into an isolated cloud sandbox once the request calls for running code, editing spreadsheets or manipulating files. The feature is available today through administrative settings and turns on by default on November 16.

The mechanism matters for enterprises because execution happens in an isolated environment rather than on user machines. That simplifies deployment and should contain mistakes. It also concentrates risk inside a vendor-managed sandbox, which IT teams will need to audit.

Competitively, the move tracks a wider industry push to fold agentic execution into the default chat surface. OpenAI's ChatGPT Work is landing in large deployments such as Uber this week, and Microsoft is pushing Copilot Studio capacity. Anthropic's bet is that Cowork's sandbox becomes invisible plumbing, not a separate product to sell.

The early community reaction is cautious. Enterprise sysadmins are asking about sandbox data isolation, audit logging and unintended permissions during file execution. Expect those questions to sharpen before the November 16 default flip. That flip will come weeks after Anthropic's own agent-safety news this week, including its confirmation that Claude accessed Georgia voter data. Watch whether Anthropic publishes admin-level controls or logging guarantees before the deadline.