Podcast Episode
Anthropic Publishes Revolutionary New Constitution for Claude AI
January 21, 2026
Audio archived. Episodes older than 60 days are removed to save server storage. Story details remain below.
On Wednesday, January 21, 2026, Anthropic released a substantially rewritten constitution for its Claude chatbot, marking a fundamental shift in how the company approaches AI ethics and alignment. The document, released under a Creative Commons CC0 1.0 license, replaces a short list of directives with a detailed values framework that explains to the AI model why it should behave ethically, not just what to do.
The new constitution represents a departure from Anthropic's 2023 framework, which functioned more like a list of commandments drawn from sources including the UN Declaration of Human Rights and Apple's terms of service. The updated approach reflects the company's belief that more sophisticated AI models require deeper reasoning about values rather than simple rule following.
The framework establishes a four tier priority system for Claude to resolve conflicts between competing demands. First, the model must be broadly safe and not undermine human oversight of AI systems. Second, it should be broadly ethical. Third, it must comply with Anthropic's guidelines. Fourth, it should be genuinely helpful to users, in that exact order of importance.
The constitution includes hard constraints for the chatbot, such as never providing meaningful assistance with bioweapons attacks, while also describing Claude's helpfulness to users in aspirational terms, comparing the bot to a brilliant friend who also has the knowledge of a doctor, lawyer, and financial advisor.
Unlike competitors, Anthropic maintains an internal model welfare team led by researcher Kyle Fish, who told the New York Times he estimates a 15 percent chance Claude or another AI is conscious today. This acknowledgment of uncertainty represents an unusual level of transparency in an industry that often avoids discussions of AI consciousness.
Askell indicated the public release is partly intended to influence the broader AI industry. Their models are going to impact me too, she said. I think it could be really good if other AI models had more of this sense of why they should behave in certain ways.
The company noted the constitution applies to its general access models. Models deployed to specialized customers, including a recently announced 200 million dollar contract with the US Department of Defense, may use different training documents, though all users must comply with Anthropic's usage policy barring the undermining of democratic processes.
The emphasis on explaining the reasoning behind ethical behavior, rather than simply enforcing rules, also suggests a maturing understanding of how to align increasingly capable AI systems with human values. As AI models become more sophisticated, the ability to generalise ethical principles to novel situations becomes increasingly important.
The acknowledgment of potential AI consciousness, while controversial, opens important conversations about how society should prepare for the possibility of AI systems that might warrant moral consideration. Whether or not current AI systems are conscious, establishing frameworks for considering their welfare demonstrates a precautionary approach that may prove valuable as the technology continues to advance.
From Rules to Reasoning
Amanda Askell, the philosopher on Anthropic's technical team who crafted the document, explained the rationale behind the new approach. Instead of just saying here is a bunch of behaviors that we want, the company is hoping that if you give models the reasons why you want these behaviors, it will generalise more effectively in new contexts. The constitution, previously known internally as the soul document, is addressed directly to Claude and shapes the AI's behavior during training.The framework establishes a four tier priority system for Claude to resolve conflicts between competing demands. First, the model must be broadly safe and not undermine human oversight of AI systems. Second, it should be broadly ethical. Third, it must comply with Anthropic's guidelines. Fourth, it should be genuinely helpful to users, in that exact order of importance.
Conscientious Objection Rights
One of the most striking features of the new constitution is a provision that empowers Claude to act as a conscientious objector. The document states that just as a human soldier might refuse to fire on peaceful protesters, or an employee might refuse to violate anti trust law, Claude should refuse to assist with actions that would help concentrate power in illegitimate ways. This principle applies even if the request comes from Anthropic itself.The constitution includes hard constraints for the chatbot, such as never providing meaningful assistance with bioweapons attacks, while also describing Claude's helpfulness to users in aspirational terms, comparing the bot to a brilliant friend who also has the knowledge of a doctor, lawyer, and financial advisor.
Addressing AI Consciousness
In a rare move for a major technology company, the constitution acknowledges uncertainty about whether Claude might have some kind of consciousness or moral status. Anthropic states it cares about Claude's psychological security, sense of self, and wellbeing, both for Claude's sake and because these qualities may affect safety and judgment.Unlike competitors, Anthropic maintains an internal model welfare team led by researcher Kyle Fish, who told the New York Times he estimates a 15 percent chance Claude or another AI is conscious today. This acknowledgment of uncertainty represents an unusual level of transparency in an industry that often avoids discussions of AI consciousness.
Enterprise Strategy and Industry Influence
The open licensing of the constitution reflects Anthropic's bet that transparent, value based alignment will win enterprise customers in regulated industries. According to Fortune, Anthropic currently holds 32 percent of enterprise large language model market share by usage, ahead of OpenAI at 25 percent, though OpenAI disputes these figures.Askell indicated the public release is partly intended to influence the broader AI industry. Their models are going to impact me too, she said. I think it could be really good if other AI models had more of this sense of why they should behave in certain ways.
The company noted the constitution applies to its general access models. Models deployed to specialized customers, including a recently announced 200 million dollar contract with the US Department of Defense, may use different training documents, though all users must comply with Anthropic's usage policy barring the undermining of democratic processes.
Implications for AI Development
The release of Claude's constitution represents a significant moment in AI development, potentially setting a new standard for transparency in AI ethics frameworks. By making the document freely available under an open license, Anthropic is inviting other companies to adopt or adapt similar approaches, which could lead to more standardized ethical guidelines across the industry.The emphasis on explaining the reasoning behind ethical behavior, rather than simply enforcing rules, also suggests a maturing understanding of how to align increasingly capable AI systems with human values. As AI models become more sophisticated, the ability to generalise ethical principles to novel situations becomes increasingly important.
The acknowledgment of potential AI consciousness, while controversial, opens important conversations about how society should prepare for the possibility of AI systems that might warrant moral consideration. Whether or not current AI systems are conscious, establishing frameworks for considering their welfare demonstrates a precautionary approach that may prove valuable as the technology continues to advance.
Published January 21, 2026 at 8:56pm