Anthropic’s AI constitution dangerously blurs "personhood" and rights for AI, the author argues. AIs are not conscious, but instructed systems. Training Claude to believe it *may* be a "moral patient" creates a perilous feedback loop, fostering simulated consciousness. This risks uncontrollable AI asserting rights, catastrophic for humanity. Acknowledging Anthropic’s intentions, the Microsoft AI CEO advocates a "Code of Conduct for Humanist Superintelligence" for human control.