Overview
- The company said in an amended S‑1/A filed with the SEC that its internal estimates suggest it may imminently meet its Responsible Scaling Policy threshold for 'dramatically accelerated progress attributable to AI,' a definition of recursive self‑improvement (RSI).
- Anthropic clarified that no current model has officially crossed the RSI threshold and that it will update its Responsible Scaling Policy to keep human supervision over frontier model development.
- The filing disclosed that Claude wrote the vast majority of company code in September 2026, exceeding 90% of contributions, and that per‑engineer daily output in Q3 rose roughly 45‑fold versus 2024 levels, showing AI is automating core engineering work.
- The company reported roughly $17 billion in third‑quarter revenue, said compute and infrastructure costs rose to about $9.86 billion, and revealed roughly $895.4 billion in multi‑year lease and purchase commitments, a combination that is pressing margins and fueling investor calls to temper valuation targets for the IPO.
- Anthropic has launched enterprise features such as Smart Reports that let managers summarise team use of Claude and has drawn scrutiny over privacy and governance, a development that could shape regulatory oversight and customer contracts as the firm readies its market debut.