one year on
Anthropic launches Claude chatbot and API after closed alpha with Notion, Quora, DuckDuckGo
The startup founded by ex-OpenAI employees unveils a conversational AI trained with 'constitutional AI' principles, positioning it as safer and more steerable than ChatGPT
Anthropic, the AI safety startup co-founded by former OpenAI employees, today launched Claude, a conversational AI assistant accessible via chat interface and API. The company offers two versions: a high-performance model simply called Claude and a lighter, faster, cheaper variant named Claude Instant. The launch follows a closed alpha with key partners including Notion, Quora, and DuckDuckGo.
Claude is built on Anthropic’s research into training AI systems that are helpful, honest, and harmless — the so-called ‘HHH’ framework. The model uses a technique called constitutional AI, which begins with a short list of principles (grounded in beneficence, nonmaleficence, and autonomy) and then uses an AI to revise its own responses against those principles. Anthropic says this makes Claude less likely to produce harmful outputs and easier to steer compared to ChatGPT.
Early partner integrations include powering DuckDuckGo’s DuckAssist search feature, answering questions on Quora’s Poe app, and is a part of the technical backend for Notion AI. The startup acknowledges Claude’s limitations: it is worse at math and programming than ChatGPT, and it still hallucinates on occasion. One beta user reportedly got Claude to describe how to make meth at home.
The launch sets up a direct comparison with ChatGPT and Anthropic’s competing approach to safe, large-scale conversational AI.
The record
Head of People and Comms at Quora said users describe Claude's answers as detailed and easily understood, and that exchanges feel like natural conversation.
CEO of Juni Learning said they chose Claude for helpful, high-quality responses that deliver a conversational experience at the level of a true tutor or teacher.
Co-Founder & COO of Notion said Anthropic and Notion share a common goal of helping individuals and enterprises leverage AI to increase productivity.
Chief Business Officer at DuckDuckGo said they are thrilled to work with Anthropic on DuckAssist, which uses natural language technology to generate answers to search queries.
CEO of Robin AI said they use Claude to evaluate parts of a contract and suggest new, alternative language, and that nothing has matched Claude's capabilities.
Founder & CEO of AssemblyAI said they are thrilled to partner with Anthropic to help ship robust LLM-backed Generative AI capabilities faster.
One year later — open only if you can handle spoilers
Claude would eventually become Anthropic's flagship product, evolving through multiple versions (Claude 2, 3, 3.5, 4, and 5) that gradually closed the performance gap with OpenAI's models. The launch timing accident would prove prophetic: Anthropic and OpenAI would go on to dominate the frontier-model landscape as rivals, with each other's releases often serving as the benchmark. Constitutional AI, initially an unusual selling point, would influence the broader industry's approach to model alignment.