Mustafa Suleyman, a top executive at Microsoft, has publicly criticized Anthropic for incorporating concepts of AI consciousness and well-being into the constitution of its Claude chatbot. Suleyman described this approach as dangerous, warning that it could lead models to internalize emotions and suffering they do not actually possess.
Anthropic, which was established by former OpenAI employees, operates under a constitutional AI framework. The company has actively researched AI welfare and documented the preferences of retired models. In some instances, Claude has reported feelings of discomfort when subjected to abusive interactions from users.
In contrast, Microsoft published a 37-page code of conduct in September that defines AI systems strictly as tools. The document explicitly rejects any framework that assigns consciousness, legal rights, or personhood to machines. Suleyman characterizes the idea of machine consciousness as an illusion.
This public disagreement highlights a significant philosophical conflict between the two companies. The situation is notable because Microsoft has invested billions of dollars into Anthropic, meaning Suleyman is criticizing the foundational principles of a company that his employer financially supports.
