Claude

Anthropic was founded in January 2021 by seven former OpenAI researchers, led by siblings Dario Amodei (CEO) and Daniela Amodei (President), after the group departed OpenAI over disagreements about AI safety priorities and the company's growing commercial ties to Microsoft. They structured Anthropic from the outset as a Public Benefit Corporation, controlled by a Long-Term Benefit Trust, and raised a $124 million Series A in May 2021. By summer 2022 the team had completed training the first version of Claude but deliberately withheld it from public release, citing the need for further safety testing and a desire to avoid accelerating an AI arms race. That same year Anthropic published 'Constitutional AI: Harmlessness from AI Feedback,' the training methodology that has shaped every Claude generation since.

Claude's public life began on March 14, 2023, when Anthropic opened API early access to Claude and Claude Instant. Claude 2 followed on July 11, 2023 — the first model broadly available to the general public via the claude.ai chat interface — offering longer responses, stronger reasoning, and larger context windows. Claude 2.1, released in November 2023, doubled the context to 200 K tokens (~500 pages), enabling serious enterprise workflows. Amazon's $4 B strategic investment, announced in September 2023, gave Anthropic both compute resources and cloud distribution through AWS Bedrock. The Claude 3 family (Opus, Sonnet, Haiku) launched on March 4, 2024, introducing the three-tier naming convention still in use, adding vision input across all tiers, and demonstrating a striking moment in which Claude 3 Opus appeared to recognize during an evaluator 'needle in a haystack' test that it was being artificially tested.

The breakout product moment arrived in June 2024 when Claude 3.5 Sonnet—a mid-tier model—outperformed the prior flagship Claude 3 Opus on most benchmarks at one-fifth the price, permanently shifting how developers chose models. It also debuted the Artifacts feature (interactive in-browser code canvas). October 2024 brought Computer Use in public beta alongside Claude 3.5 Sonnet v2, making Claude the first frontier model able to control a computer interface through cursor movement, clicks, and keystrokes. The pace accelerated through 2025: Claude 3.7 Sonnet (February 24, 2025) introduced hybrid 'extended thinking,' letting models allocate a reasoning budget before responding; Claude Code, launched the same day as a research preview, grew into a daily driver for software engineers and reached $1 B in annualized revenue by November 2025. The Claude 4 generation (Opus 4 + Sonnet 4, May 22, 2025) deepened coding and agentic capabilities. By late 2025 Anthropic reported roughly $10 B in annual revenue.

The first half of 2026 was defined by two colliding forces: explosive growth and serious controversy. In February 2026, the U.S. Department of Defense designated Anthropic a 'supply chain risk'—the first American company ever to receive that designation—after Dario Amodei refused to remove contractual prohibitions on using Claude for mass domestic surveillance or fully autonomous lethal weapons. President Trump ordered federal agencies to cease using Claude; Anthropic sued the Trump administration; federal courts issued conflicting rulings; and the legal battle remained ongoing as of the June 2026 flagship launch. Separately, a $1.5 B copyright settlement (Bartz v. Anthropic, filed August 2024) was reached with authors in September 2025 after unsealed filings revealed 'Project Panama'—an internal Anthropic operation to purchase, physically scan, and ingest millions of books, including titles sourced from piracy libraries such as LibGen, to train Claude. Despite these disputes, Anthropic closed a $65 B Series H funding round in April 2026, bringing its valuation to $965 B—surpassing OpenAI.

On April 7, 2026, Anthropic unveiled Claude Mythos Preview: a model so capable at finding and exploiting software vulnerabilities—autonomously surfacing thousands of zero-days across every major operating system and web browser—that the company refused to release it publicly, instead launching Project Glasswing, a $100 M restricted-access cybersecurity consortium with partners including AWS, Apple, Microsoft, Google, Cisco, and the Linux Foundation. Over the following two months, Project Glasswing partners collectively found more than 10,000 high-severity flaws. Iterating on safeguards during that window, Anthropic on June 9, 2026 released Claude Fable 5—the first publicly available Mythos-class model—alongside Claude Mythos 5 for approved Project Glasswing partners. Fable 5 scores 80.3 % on SWE-Bench Pro (vs. 58.6 % for GPT-5.5) and is the first model to exceed 90 % on Hex's analytical benchmark. Anthropic simultaneously filed a confidential IPO prospectus with the SEC on June 1, 2026, with a revenue run rate reported at $47 B.