Archive · 22 Jan 2026
Anthropic drafts constitution for Claude AI models
Anthropic, a company that makes AI, has created a special set of rules called a 'constitution' for their AI named Claude.
This constitution is like a guidebook that tells Claude how to behave.
It includes rules about being safe, honest, and helpful.
Anthropic wants Claude to follow these rules to make sure it acts in a good way.
They got help from experts in different fields to make this constitution.
The rules will be used to train Claude and make sure it understands how to act properly.
However, Anthropic knows that Claude might not always follow the rules perfectly, but they believe this constitution will help guide Claude to be a good and helpful AI.
Anthropic published a 'constitution for Claude' to guide the AI's behavior and ethics.
The constitution outlines principles of safety, ethics, compliance, and helpfulness.
It will be used to train general-purpose Claude models, excluding specialized ones.
Anthropic acknowledges that Claude might not always adhere to the constitution's ideals.
The constitution was drafted with feedback from experts in various fields.
- Who
- Anthropic
- What
- Drafted a constitution for Claude AI models
- When
- January 22, 2024
- Where
- Not specified
- Why
- To define AI behavior and ethics, and align with human values
- Organization
- Anthropic
- AI Model
- Claude
- Document Name
- Constitution for Claude
- Purpose
- Define AI behavior and ethics
- Expert Feedback
- Law, philosophy, theology, psychology
- Principles
- Safety, ethics, compliance, helpfulness
- Applicability
- General-purpose Claude models only
- Date Published
- January 22, 2024
