Author Rejects Anthropic's 84-Page Claude Constitution, Says AI Lacks Consciousness and Moral Agency
Updated
Updated · The Atlantic · Jun 3
Author Rejects Anthropic's 84-Page Claude Constitution, Says AI Lacks Consciousness and Moral Agency
2 articles · Updated · The Atlantic · Jun 3
Summary
Anthropic’s 84-page “constitution” for Claude is attacked as anthropomorphic marketing, with the author arguing that large language models are predictive text systems, not conscious beings or moral agents.
LLM chat feels human because users co-author a role-play transcript one word at a time, the essay says; fluent dialogue is a text deepfake, not evidence of feelings, understanding or subjective experience.
That distinction matters because treating Claude as morally aware risks shifting responsibility from companies and users onto software that cannot bear legal, social or ethical accountability for its actions.
The essay argues moral reasoning requires embodiment, emotions and lived consequences, making Claude’s first-person claims such as “I understand” fundamentally misleading even if its outputs can mimic principled speech.
Anthropic’s framing is portrayed as commercially useful but conceptually incoherent: if Claude were truly conscious, the company would face far heavier duties—potentially including liability, protections or even reparations.
As AI 'method actors' learn our emotions, are we building safer tools or just more sophisticated psychological traps?
With AI agents now making business decisions, who is legally responsible when these non-conscious tools inevitably fail?
If today's AI is a dead end for consciousness, what is the evolutionary path to building genuinely sentient machines?
The Claude Constitution (2026): How Anthropic’s 84-Page AI Charter Redefines Ethics, Consciousness, and Corporate Power
Overview
In January 2026, Anthropic released the Claude Constitution, marking a major step forward in AI governance. Evolving from its 2023 version, this new constitution sets a benchmark for ethical and responsible AI design. It uses a philosophical, virtue-ethics approach to guide Claude’s behavior, focusing on principles that encourage positive outcomes. The framework is built on a clear principal hierarchy, ensuring layered accountability and control. Explicit hard constraints are included to set clear boundaries for Claude’s actions, helping prevent harmful outputs. Together, these elements create a robust foundation for trustworthy and safe AI operation.