Anthropic

Anthropic is an artificial intelligence company founded in 2021 by former OpenAI employees, including siblings Dario and Daniela Amodei. Dario serves as chief executive, while Daniela is president. The company is based in San Francisco and describes its mission as building reliable, interpretable, and steerable AI systems. Before starting Anthropic, Dario worked on AI research at OpenAI, where Daniela also held a leadership role in safety and policy.

Anthropic introduced Claude, its conversational AI assistant, in March 2023. The company subsequently released Claude 2 in July 2023, followed by the Claude 3 family in March 2024. That family comprised Haiku, Sonnet, and Opus, models positioned at different points on the speed and capability spectrum. In June 2024, Anthropic launched Claude 3.5 Sonnet, highlighting improvements in coding and visual reasoning. Claude is accessible through a website, mobile apps, and an API for developers; Anthropic also offers enterprise access for organizations deploying its models in their workflows.

A distinctive part of Anthropic’s research is Constitutional AI, a method described in a 2022 paper. Instead of relying solely on human ratings, the approach uses a written set of principles to guide model critiques and revisions of responses. Anthropic has published research on interpretability, including work probing how concepts are represented inside neural networks. Its Responsible Scaling Policy, first published in 2023, sets out safeguards tied to assessments of increasingly powerful models. These projects connect the firm’s product development to its stated concerns about misuse and loss of control.

Anthropic has attracted substantial investment from Amazon and Google. In 2023, Amazon announced a strategic partnership with Anthropic, including plans to invest up to $4 billion and make Anthropic models available through Amazon Bedrock. Anthropic agreed to use Amazon Web Services as a primary cloud provider. Its models are also available through Google Cloud’s Vertex AI platform. These arrangements have put Claude into cloud services used by businesses building applications on top of models.

In October 2024, Anthropic introduced a beta computer use capability for Claude 3.5 Sonnet, allowing developers to let the model interact with on-screen interfaces by moving a cursor, clicking, and typing. Anthropic cautioned that the feature was experimental and could make mistakes. The release illustrated a shift from answering prompts toward carrying out multistep tasks in software, while keeping safety questions central to its announcements.