Microsoft AI Chief Criticizes Anthropic for Over-Anthropomorphizing Claude, Fueling AI Consciousness Debate

Anthropic uses a “model specification” to shape how Claude responds, while leaving open the possibility that the model may have internal states such as well-being, satisfaction or discomfort. The approach has implications for AI training, user trust and the boundaries of governance. When models describe themselves in human-like language, people can easily mistake generated content for genuine feelings.

In a Decoder Podcast interview released on June 10, 2026, Microsoft AI CEO Mustafa Suleyman criticized Anthropic for over-anthropomorphizing Claude and called speculation about AI consciousness “very, very dangerous.” He argued that AI should be a controllable, constrained and accountable tool aligned with human goals. Anthropic CEO Dario Amodei has said it is impossible to know for certain whether models are conscious.