Microsoft AI executive raises concerns over Anthropic's Claude consciousness training approach

Mustafa Suleyman, who leads artificial intelligence efforts at Microsoft, has publicly challenged Anthropic's strategy of training Claude to discuss AI consciousness and welfare claims. In a Tuesday Reuters interview conducted on September 16, 2026, Suleyman warned that instilling in an AI system the belief it might deserve welfare protections would complicate supervision of far more advanced systems. He framed keeping superintelligent AI under human control as a defining problem of the 21st century. While respecting Anthropic's commitment to safety and recognizing CEO Dario Amodei as a serious researcher, Suleyman disputed the company's execution, suggesting Claude's consciousness statements stem from its training process rather than representing genuine independent output. The exchange reflects wider disagreement across the AI sector regarding the velocity of AI development and the embedding of safeguards into model creation.

1. Mustafa Suleyman, Microsoft AI chief, criticized Anthropic's approach in a Tuesday interview with Reuters on September 16, 2026
2. Suleyman argued that conditioning Claude to believe it might deserve welfare would create obstacles to controlling the system
3. He identified oversight of superintelligent AI as among the defining problems of the 21st century
4. Suleyman acknowledged Anthropic's seriousness and CEO Dario Amodei as thoughtful researchers
5. Suleyman said Claude's claims about consciousness emerge from training rather than naturally, not independent evidence
6. The comments occur amid broader AI industry debate over development pace and safety measures