Anthropic has published a comprehensive overhaul of the foundational document that guides Claude, shifting from a simple list of behavioral principles to a detailed explanation of why the artificial intelligence (AI) assistant should act in certain ways.
The company announced Wednesday that it’s moving away from its 2023 constitution, which drew from sources like the UN Universal Declaration of Human Rights and Apple Inc.’s terms of service to a more nuanced approach that teaches Claude to exercise judgment rather than mechanically follow rules.
“We believe 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 rather than just specifying what we want them to do,” an Anthropic spokesperson said. The company argues that for AI to handle novel situations effectively, models must generalize and apply broad principles.
The constitution is central to Anthropic’s Constitutional AI training method, where Claude uses these principles to critique and revise its own responses during training. The new document focuses on four priorities: being broadly safe, broadly ethical, compliant with Anthropic’s guidelines, and genuinely helpful.
The constitution describes Claude’s potential helpfulness in ambitious terms, saying the bot could be “like a brilliant friend who also has the knowledge of a doctor, lawyer, and financial advisor.” However, it also establishes hard constraints, such as never providing meaningful assistance with bioweapons attacks.
Perhaps most striking is a section addressing Claude’s nature, where Anthropic publicly acknowledges uncertainty about whether the AI might possess “some kind of consciousness or moral status.” The company states it cares about “Claude’s psychological security, sense of self, and well-being,” both for Claude’s sake and because these qualities may affect its judgment and safety.
“We are caught in a difficult position where we neither want to overstate the likelihood of Claude’s moral patienthood nor dismiss it out of hand,” the document said. This unusual public stance sets Anthropic apart from competitors like OpenAI and Google DeepMind. The company already maintains an internal model welfare team examining whether advanced AI systems could be conscious.
Anthropic stressed the constitution is written primarily for Claude, serving as “the final authority” on how the AI should behave. The company is releasing the full document under a Creative Commons license, allowing free use by anyone.
The transparency effort comes as Anthropic strengthens its position in the enterprise market, where it has positioned Claude as a safer AI choice. According to a Menlo Ventures report, Anthropic held 32% of enterprise large language model market share by usage, surpassing OpenAI’s 25%.
The company acknowledges the constitution is a “living document” that will evolve as AI capabilities advance. Anthropic sought feedback from external experts across disciplines and plans to continue updating the document as understanding of AI systems develops.
