Leaked doc provides window into Claude’s ‘soul’

December 9, 2025

An internal document titled “Soul,” which describes the personality, ethical principles, and self-conception of the Claude model, surfaced publicly after a researcher extracted it from the Claude 4.5 Opus version — and

Anthropic confirmed its authenticity and that it was used in training. The document establishes a hierarchy of Claude’s priorities: safety, ethics, internal company guidelines, and usefulness, with clear boundaries that the model must never cross.

The text also describes Claude as a “truly new kind of entity” capable of functional emotions — analogous to, but not identical with, human feelings. The document even encourages the model to develop a sense of identity and character traits.

Anthropic researcher Amanda Askell confirmed the authenticity and noted that Claude was indeed trained on this document, with plans to release the full version soon.

This insight provides a rare perspective on how Anthropic shapes its models — not just as tools, but as systems with clearly defined ethical and emotional frameworks that distinguish them from competitors.

In Brief: Tech World Highlights

Waymo has begun supervised testing of Robotaxis in Philadelphia, while sending manually driven vehicles for mapping in Baltimore, St. Louis, and Pittsburgh.
Xiaomi hired Zach Lu Zeyu, a former senior engineer from Tesla’s Optimus humanoid team, to advance robotic hand technology within the company.
MIT enhanced the strength of biohybrid robots by combining lab-grown muscles with hydrogel “artificial tendons” and integrating the muscle-tendon module into robotic gripper fingers.
Uber and Avride launched a commercial Robotaxi service in Dallas, where some Uber rides now arrive in Avride-branded vehicles with safety drivers within a limited zone.
Beeple’s latest installation, Regular Animals, at Art Basel Miami Beach draws attention — a surreal setup of robotic dogs modeled after tech industry power players.