Claude Advanced Tool Use

Claude Opus 4.5

Effective harnesses for long-running agents

Disrupting the first reported AI-orchestrated cyber espionage campaign

Code execution with MCP: Building more efficient agents

Measuring political bias in Claude

Microsoft, Nvidia and Anthropic announce strategic partnerships

Anthropic invests $50B in US AI infrastructure

Claude Haiku 4.5 System Card [pdf]

A small number of samples can poison LLMs of any size