How does an AI know what it’s supposed to do and not do? Their creators can tell the public all about their ideals, but if the AI itself doesn’t know what its guardrails are, how can the creator be sure it acts in alignment with its values? Anthropic, creator of Claude AI that competes with ChatGPT and Google Gemini, released Claude’s Constitution on January 21, 2026.
“As AI models become more integrated in society, it’s important people understand how AI models are built,” an Anthropic spokesperson. “This document is our attempt at that as a crucial part of our training process.”
For my dearest Claude…
What sets this document apart from those that Anthropic has published before, which were public-facing and meant to declare to the people their company’s values, the Constitution is designed for Claude itself.
“The constitution is written for Claude,” said Anthropic. “It explains what Claude is, the context in which it operates, and the kind of entity we’d like it to be. We believe that in order to be safe and helpful, AI models like Claude need to understand why we want them to behave in certain ways, rather than simply being told what to do.”
Anthropic had previously had a Constitution for Claude, but they rewrote it to the version released this week. “Our previous Constitution was composed of a list of standalone principles. We’ve come to believe that a different approach is necessary. We think that in order to be good actors in the world, AI models like Claude need to understand why we want them to behave in certain ways, and we need to explain this to them rather than merely specify what we want them to do.”
Claude’s Constitution was written primarily by Amanda Askell, Claude’s in-house philosopher, whom Anthropic called responsible for shaping Claude’s character. Joe Carlsmith, Chris Olah, Jared Kaplan, and Holden Karnofsky also contributed to the document.
The post Claude’s Creator Wrote a Constitution for Claude appeared first on VICE.
