Anthropic is an American AI safety company founded in 2021 by Dario Amodei (CEO) and Daniela Amodei (President), both former executives at OpenAI. Structured as a Public Benefit Corporation, Anthropic develops the Claude family of AI assistants and pioneered Constitutional AI, a methodology for training AI systems to be helpful, harmless, and honest. By early 2026, Anthropic reached a $380 billion post-money valuation with $14 billion in annualized revenue, holding 40% of the enterprise LLM market.1)
Dario Amodei served as VP of Research at OpenAI, and Daniela Amodei led Safety and Policy there before the siblings departed over disagreements regarding AI safety priorities and scaling approaches. Anthropic was established with a core thesis that increasingly powerful AI systems require safety research to be deeply integrated into model development rather than applied as an afterthought.2)
Released in early 2024, the Claude 3 lineup introduced a tiered approach:
Iterative improvements delivering substantially better performance across benchmarks while maintaining the safety properties of the Claude 3 family.
The latest generation extends Claude into agentic capabilities. Claude Code became Anthropic's breakout product, generating $1 billion in revenue within six months for coding and engineering workflows. Cowork expanded capabilities to knowledge work including sales, legal, and finance through eleven open-source plugins.3)
Constitutional AI (CAI) is Anthropic's core training methodology. Instead of relying solely on human feedback (RLHF), CAI trains models against a set of explicit principles (a “constitution”) that define acceptable behavior. The model critiques and revises its own outputs according to these principles, reducing harmful outputs while maintaining helpfulness. This approach enables safer deployment in regulated industries and underpins Anthropic's enterprise trust model.4)
Anthropic publishes detailed model cards, conducts red-teaming exercises, and invests heavily in interpretability research. In March 2026, the company launched The Anthropic Institute, led by co-founder Jack Clark, to research AI's societal impacts, economic effects, and safety methodologies, sharing findings publicly.5)
Anthropic generates 85% of its revenue from enterprise customers (compared to OpenAI's 85% from consumer subscriptions). Key enterprise metrics:
| Period | Revenue (Annualized) |
|---|---|
| End 2024 | $1B |
| Q1 2025 | $2B |
| May 2025 | $3B |
| August 2025 | $5B+ |
| 2025 Full Year | ~$9B |
| 2026 Target | $18-26B |
The February 2026 Series G raised $30 billion led by GIC and Coatue, with participation from Microsoft, NVIDIA, and Sequoia, at a $380 billion post-money valuation. Anthropic projects positive cash flow by 2027-2028.7)
Anthropic has a deep strategic partnership with Amazon Web Services (AWS), distributing Claude through AWS Bedrock. Claude is also available on Google Cloud Vertex AI and Microsoft Azure, providing broad enterprise reach.