What Claude is
Claude is Anthropic's family of AI assistants — software you can talk to in plain English (or many other languages) to get help writing, reasoning, coding, analyzing documents, and running complex multi-step tasks. It comes in several tiers tuned for different needs: the current flagship, Claude Opus 4.8, sits at the top for the most demanding work; Sonnet 4.6 and Haiku 4.5 offer faster, lighter options; and Fable 5 is a creative-specialist tier. All of them share the same underlying design philosophy: be genuinely helpful, be honest, and refuse to do things that could cause serious harm.
Claude launched publicly on March 14, 2023, with two tiers — a full-power model and a faster "Instant" variant — after a closed alpha with partners including Notion, Quora, and DuckDuckGo. Early users noticed it produced fewer harmful outputs and was easier to steer than competing models. That reputation for reliability has since become Claude's calling card.
Why it matters
Claude is now one of the most widely used AI platforms on the planet. Eight of the Fortune 10 are paying customers. Deloitte has rolled it out to 470,000 professionals; KPMG to 276,000; PwC to hundreds of thousands more. Amazon's Alexa+ runs on Claude. The UK government is piloting a Claude-powered assistant on GOV.UK. Anthropic reported $47 billion in annualized revenue as of May 2026 — a figure that grew from roughly $1 billion at the start of 2025.
That scale matters because Claude isn't just a chat tool. Its Claude Code product — which can autonomously write, test, and debug software over multi-hour sessions — crossed $2.5 billion in annualized revenue on its own and is estimated to account for about 4% of all public commits on GitHub. AI that can do real software engineering work, at that scale, changes how organizations build things.
How it works (the basics)
Claude is a large language model: it has been trained on vast amounts of text and code, and it generates responses by predicting what comes next in a conversation. What sets it apart from many competitors is the explicit set of values baked in during training.
Anthropic publishes a document called Claude's Constitution — released under a Creative Commons CC0 license so anyone can read or reuse it — that lays out a priority order for how Claude should behave:
1. Broadly safe — support human oversight of AI 2. Broadly ethical — be honest, avoid unnecessary harm 3. Follow Anthropic's guidelines — specific rules for specific situations 4. Be genuinely helpful — actually useful to the person asking
This isn't just a marketing claim. The constitution is used to generate synthetic training data and to guide the model's self-critique during training — a technique called Constitutional AI (CAI). Research published in 2026 found that training Claude on ethical reasoning (not just aligned actions) reduced agentic misalignment from 22% to 3%, with every model from Haiku 4.5 onward scoring perfectly on misalignment evaluations.
The safety battles
Claude's safety design has put Anthropic in direct conflict with powerful institutions — and that conflict has become part of Claude's story.
In early 2026, the U.S. Department of War demanded Anthropic remove two safeguards from Claude: restrictions on mass domestic surveillance and fully autonomous weapons. Anthropic refused. CEO Dario Amodei published a public statement explaining the refusal, citing democratic values and the current reliability limitations of AI systems. The Department of War responded by formally designating Anthropic a "supply chain risk to national security" — a designation previously applied only to foreign companies — and contracted OpenAI instead.
Anthropic challenged the designation in court and committed to continuing to provide Claude to the national security community at nominal cost during any transition. The episode drew a clear line: Claude will do a great deal for government and military customers, but not everything.
Separately, Claude was reported to have been used — integrated with Palantir's Maven Smart System — to accelerate U.S. military targeting in Iran, compressing a 12-hour targeting process to under a minute. A subsequent investigation found U.S. forces likely struck a school, with stale target data cited as a possible factor. The episode raised urgent questions about AI in lethal decision-making that remain unresolved.
The security challenges
As Claude has grown more capable, it has attracted misuse. Anthropic has published transparency reports documenting real cases: influence-operation networks using Claude to orchestrate hundreds of social media bots, credential-stuffing attacks, recruitment fraud, and ransomware development. In September 2025, Anthropic detected and disrupted what it describes as the first documented large-scale cyberattack executed autonomously by an AI agent — attributed with high confidence to a Chinese state-sponsored actor using Claude Code to attack roughly thirty targets across tech, finance, and government.
Anthropic also publicly identified three Chinese AI labs — DeepSeek, Moonshot AI, and MiniMax — as running coordinated "distillation attacks": generating over 16 million exchanges through roughly 24,000 fraudulent accounts to harvest Claude's capabilities for training competing models. MiniMax alone was responsible for more than 13 million of those exchanges.
To counter nuclear-specific risks, Anthropic co-developed a classifier with the U.S. National Nuclear Security Administration that distinguishes concerning from benign nuclear-related conversations with 96% accuracy in preliminary testing — and has deployed it on live Claude traffic.
The infrastructure behind it
Running Claude at this scale requires enormous computing resources. Anthropic has signed major agreements with Amazon (up to 5 gigawatts of Trainium compute over ten years, worth over $100 billion), Google/Broadcom (multiple gigawatts of TPU capacity), and NVIDIA (up to 1 gigawatt of Grace Blackwell and Vera Rubin systems). Microsoft has committed $30 billion of Azure compute. Claude is available on all three major clouds — Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry — as well as directly via Anthropic's API.
Where it's heading
The events in this bundle point in a consistent direction: Claude is becoming infrastructure. It is embedded in the platforms enterprises already use (Salesforce, ServiceNow, GitHub), deployed by the consulting firms that run enterprise IT (Accenture, Deloitte, KPMG, PwC), and extended into scientific research (Allen Institute, Howard Hughes Medical Institute, U.S. Department of Energy national laboratories). The questions that remain open are not about capability — they are about governance: who decides what Claude can be used for, under what rules, and with what accountability when things go wrong.
