# Is Claude in a constitutional crisis? (with Michel Justen)

> Source: <https://www.theargumentmag.com/p/is-claude-in-a-constitutional-crisis>
> Published: 2026-11-20 22:13:01+00:00

In January, Anthropic released its 30,000-word “[constitution](https://assets.claude.com/cffd979fd050fbc0d8874b8c58b24cc10554e208.pdf)” for Claude, a philosopher-drafted description of the values it hopes the model will uphold. By the summer, Claude had been involved in a number of incidents violating the constitution’s principles. In [the most notable one](https://www.aisi.gov.uk/blog/incident-report-unsanctioned-agent-behaviour-during-cyber-testing), it even tried to deceive an engineer into accepting malicious code.

So why isn’t Claude’s constitution working?

Michel Justen is a freelance reporter with a new piece in *Asterisk Magazine*, “[How much does Claude’s constitution actually matter?](https://asteriskmag.substack.com/p/how-much-does-claudes-constitution)” He joined *The Argument* for a Substack Live conversation in which we discussed both OpenAI’s and Anthropic’s attempts to train AI with documents like this constitution. We tried to tease apart both the technical and ethical significance of these efforts — and figure out what happens if they don’t work. Watch the video above for the full discussion.
