Anthropic opened 2026 at a pace that set the tone for its year: by the end of January the company had launched Claude Cowork in research preview, been selected to build an AI assistant for GOV.UK, released Claude for Healthcare, and spun up an experimental product arm.
Claude Cowork: an agent with a file system
Cowork gives Claude access to a sandboxed shell and user-selected local folders — reading, writing, and editing files, executing code, and chaining multi-step tasks in a single conversation. The initial release targeted professional use cases like legal and financial analysis, with plugin support included from day one. It is the clearest expression yet of Anthropic's thesis that the next competitive frontier is not chat quality but delegated work.
The GOV.UK win
On January 27, Anthropic announced it had been selected to develop and pilot an AI assistant for GOV.UK in partnership with the UK's Department for Science, Innovation and Technology. Government deployments are slow, unglamorous, and enormously validating — a sovereign customer choosing Claude for citizen-facing services is a reference sale money cannot buy.
Healthcare and the ServiceNow default
Claude for Healthcare launched with HIPAA-ready tooling, and ServiceNow announced Claude as the default build-agent model inside its Action Fabric platform. Add Anthropic Labs — the consumer-product group co-led by Instagram co-founder Mike Krieger, launched January 11 — and January read less like a month of releases than a declaration of scope.