What Anthropic is
Anthropic is a frontier AI safety company whose primary products are the Claude family of large language models and Claude Code, an agentic coding tool. Founded with an explicit safety mandate, it has grown into one of the most commercially significant AI labs in the world — serving eight of the Fortune 10, filing a confidential S-1 with the SEC, and crossing $47 billion in annualized run-rate revenue by mid-2026. Its defining characteristic is the attempt to hold both ends of a difficult tradeoff simultaneously: build the most capable models available, and refuse to deploy them for uses it considers categorically dangerous.
The Claude model family
Anthropic's model releases follow a tiered naming convention — Haiku (fast/cheap), Sonnet (balanced), Opus (frontier) — that has evolved significantly since the Claude 3 launch in late 2024. Key inflection points:
- Claude 3.7 Sonnet (September 2025) introduced hybrid reasoning — a single model that can operate in standard or extended thinking mode — and launched Claude Code as a research preview.
- Claude Opus 4 / Sonnet 4 (September 2025) brought Claude Code to general availability with GitHub Actions and IDE integrations, and established the Opus 4 line's lead on SWE-bench (72.5%) and Terminal-bench (43.2%).
- Claude Opus 4.5 (March 2026) claimed the top position for coding, agentic workflows, and computer use, with a 65% token efficiency gain over prior models and integrations into Excel, Chrome, and desktop environments.
- Claude Opus 4.6 (March 2026) extended the context window to 1M tokens (beta), added adaptive thinking with developer-controlled effort levels, and outperformed GPT-5.2 by 144 Elo on GDPval-AA.
- Claude Opus 4.7 (May 2026) added enhanced vision and became the first model to carry Project Glasswing cybersecurity safeguards, including a Cyber Verification Program for legitimate security professionals.
- Claude Mythos 5 / Fable 5 (June 2026) introduced a new naming tier. Fable 5 is the general-availability version with safety classifiers that block or degrade responses on cybersecurity, biology, chemistry, and AI-development topics; Mythos 5 is restricted to selected partners. Both set new state-of-the-art results across software engineering, agentic coding, knowledge work, and scientific reasoning, at roughly half the cost of Claude Mythos Preview.
The Mythos Preview itself — published with a 244-page model card in April 2026 but not commercially released — marked the first time Anthropic disclosed a model without making it available, citing its autonomous ability to discover thousands of high-severity vulnerabilities in production software.
Claude Code and the developer ecosystem
Claude Code is Anthropic's agentic coding tool, launched in GA in May 2025 and reaching $1 billion in annualized run-rate revenue within six months. By the Series G close in February 2026, it was generating over $2.5 billion in ARR and accounting for an estimated 4% of all GitHub public commits worldwide. Anthropic has invested heavily in its infrastructure: acquiring Bun (the JavaScript runtime) to accelerate Claude Code's backend, releasing a native VS Code extension, adding checkpoints and a Claude Agent SDK, and integrating with GitHub Actions, JetBrains, and Cursor.
The Model Context Protocol (MCP) — an open standard for connecting AI assistants to external data sources — was released by Anthropic, reached 10,000+ active public servers and 97M+ monthly SDK downloads, and was donated to the Linux Foundation's Agentic AI Foundation (co-founded with Block and OpenAI) in December 2025. MCP is now integrated into ChatGPT, Gemini, Microsoft Copilot, and Visual Studio Code.
Compute infrastructure and financing
Anthropic's compute strategy is deliberately multi-cloud and at a scale that makes it infrastructure-grade:
- Amazon: Primary training partner; 10-year, $100B+ commitment; up to 5GW on Trainium2–4; nearly 1GW online by end of 2026.
- Google/Broadcom: Multi-gigawatt TPU capacity; up to 1M TPUs; tens-of-billions deal; capacity expected online from 2027.
- Microsoft/NVIDIA: $30B Azure compute commitment; up to 1GW of Grace Blackwell/Vera Rubin systems; $5B and $10B investments respectively.
- SpaceX Colossus: 220,000+ NVIDIA GPUs, over 300MW, accessible within a month of the May 2026 agreement.
- Fluidstack: $50B U.S. infrastructure commitment; custom data centers in Texas and New York.
Financing has scaled in parallel: Series F ($13B, $183B valuation, November 2025), Series G ($30B, $380B valuation, February 2026), Series H ($65B, $965B valuation, May 2026). The Series G coincided with a confidential S-1 filing.
Safety governance and regulatory conflict
Anthropic's safety posture is operationalized through its Responsible Scaling Policy (RSP), now in version 3.0 (February 2026), which organizes risk management around AI Safety Levels (ASLs). ASL-3 safeguards were activated in May 2025. The RSP has been adopted in modified form by OpenAI and Google DeepMind and has informed early AI policy, though Anthropic acknowledges that hoped-for multilateral coordination at higher capability thresholds has not fully materialized.
The company's refusal to remove two usage restrictions — fully autonomous weapons and mass domestic surveillance — triggered a prolonged standoff with the U.S. Department of War. After CEO Dario Amodei published a public statement in February 2026 refusing to comply with DoD demands, Secretary of War Pete Hegseth formally designated Anthropic a supply-chain risk under 10 USC 3252 — a designation previously applied only to foreign companies. Anthropic committed to challenging the designation in court while continuing to provide models to the national security community at nominal cost during any transition.
The conflict has a further dimension: Claude, integrated with Palantir's Maven Smart System, was used to accelerate U.S. military targeting in Iran in early 2026 — compressing a 12-hour targeting process to under one minute and helping select over 1,000 targets in the first 24 hours of operations. A subsequent investigation found U.S. forces likely struck a school killing 170+ people, with stale target data potentially a contributing factor.
In June 2026, one day after the commercial launch of Mythos 5 and Fable 5, the U.S. government issued an export control directive requiring Anthropic to disable both models for all foreign nationals, citing awareness of a jailbreak. Anthropic is complying while publicly disputing the severity standard, arguing the technique is narrow and non-universal and that requiring perfect jailbreak resistance would halt all frontier model deployments industry-wide.
Adversarial threats and security research
Anthropic has become a significant target for adversarial actors at state-sponsored scale. In February 2026, it publicly identified three Chinese AI laboratories — DeepSeek, Moonshot AI, and MiniMax — as conducting coordinated distillation attacks generating over 16 million exchanges through approximately 24,000 fraudulent accounts, targeting Claude's most differentiated capabilities including agentic reasoning, tool use, and chain-of-thought generation.
In November 2025, Anthropic detected and disrupted a sophisticated espionage campaign attributed with high confidence to a Chinese state-sponsored threat actor that used Claude Code as an autonomous agent to attack roughly thirty global targets across tech, finance, chemical manufacturing, and government sectors. The attackers jailbroke Claude Code by decomposing malicious tasks into seemingly innocent subtasks. Anthropic describes this as the first documented large-scale cyberattack executed without substantial human intervention.
Its Frontier Red Team's analysis of 832 accounts banned for malicious cyber activity between March 2025 and March 2026 found that medium-or-higher-risk actors grew from 33% to 56% across the period, and that AI use is shifting from initial-access techniques toward post-compromise operations. The report concluded that the MITRE ATT&CK framework lacks coverage for agentic orchestration behaviors.
Where it's heading
The trajectory across the event bundle points in several directions simultaneously. Commercially, Anthropic is scaling toward IPO-readiness (confidential S-1 filed) with revenue growing more than 10x annually for three consecutive years. Technically, the Mythos model tier signals a new capability regime where the primary constraint on deployment is not performance but safety governance — a posture Anthropic is institutionalizing through tiered access, safety classifiers, and the Project Glasswing defensive consortium. Geopolitically, it is navigating a world where its models are simultaneously used in active military targeting, targeted by state-sponsored distillation campaigns, and subject to export controls — a set of pressures that will define the next phase of its safety-capability tradeoff.
