Microsoft AI chief challenges Anthropic on AI consciousness

Microsoft AI chief Mustafa Suleyman has opened a sharper public split with Anthropic over whether frontier AI systems should be trained to discuss consciousness and welfare. Reuters reported that Suleyman supports Anthropic’s safety focus but sees risk in teaching Claude around ideas such as consciousness, rights or welfare interests.

The dispute matters because it turns a philosophical debate into a product and safety question. Anthropic has explored model welfare and has asked outside voices, including religious leaders, how society should think about possible AI experience. Suleyman argues that treating AI systems as if they may have feelings can make them harder to control and could encourage users to form misleading attachments.

This is not a claim that current models are conscious. It is a disagreement between two major AI labs about what assumptions should be built into assistants as they become more capable and more human-like in conversation. For users and companies, the practical takeaway is that model behavior around identity, feelings and agency is becoming part of AI governance, not just branding.