Anthropic

The inception of Anthropic followed a significant internal shift within /openai.html during late 2020 and early 2021. Dario Amodei, who had served as the Vice President of Research, led a group of seven researchers in departing the organization. This collective included his sister, Daniela Amodei, and researcher Chris Olah.

The group's exit was reportedly motivated by divergent views on the commercial direction and safety protocols of their previous employer. They sought to establish a research environment where the predictability and interpretability of large-scale systems took precedence over rapid deployment.

This ethos led to the formation of the topic as a public benefit corporation based in San Francisco, /california.html. The founders intended to prioritize the development of reliable and steerable machine learning systems through a safety-first methodology.

During its first year, the team focused on "Constitutional" frameworks for training. This methodology aimed to provide systems with a written set of principles to guide their behavior, rather than relying solely on human feedback. The objective was to create models that remained aligned with human values.

Financial support for the venture materialized through a substantial Series A funding round. In April 2022, the organization announced it had secured $580 million in capital. This round was notable for its scale and the profile of its primary contributors.

The majority of this initial funding, approximately $500 million, originated from FTX under the leadership of Sam Bankman-Fried. Other participants in the early financing stages included Jared McCaleb and Dustin Moskovitz. These resources allowed the company to expand its technical infrastructure.

By the summer of 2022, the research team completed the training of the first iteration of its primary model, Claude. Despite the technical milestone, the leadership elected to withhold a public release. They cited a requirement for further internal safety evaluations.

This decision reflected a strategic desire to avoid accelerating competitive pressures in the technology sector. The organization maintained a focus on "mechanistic interpretability," a field dedicated to understanding the internal workings of neural networks. This research sought to move toward more transparent systems.

Early documentation from the period emphasizes the importance of "scaling laws." The founders observed that as systems grew larger, their capabilities and risks increased in predictable patterns. This observation informed their cautious approach to deployment and development.