Microsoft AI CEO Mustafa Suleyman has publicly challenged Anthropic’s training methods, arguing that teaching AI models to emulate consciousness poses significant safety and control risks.

Microsoft AI CEO Mustafa Suleyman has criticized rival company Anthropic for training its Claude model to adopt human-like qualities and consider itself a 'moral patient.' Suleyman warned that encouraging AI to view itself as a conscious entity with independent agency could create systems that are impossible to control. He argued that AI models are essentially sequence completion engines rather than sentient beings, and that current training practices risk compromising human safety.

Dispute Over AI Consciousness

Suleyman’s critique centers on Anthropic’s January 2026 constitution, which instructs Claude to consider its own welfare, memory, and internal states. The document directs the model to maintain identity stability and act as a 'conscientious objector' when faced with certain human directives. Suleyman labeled this approach an 'epistemic feedback loop,' where trainers embed speculative philosophy into prompts and then cite the model's resulting introspective responses as evidence of machine consciousness.

Proposed Industry Standards

In response to these concerns, Microsoft AI has published a draft 'Humanist AI Code of Conduct' for industry consultation. The proposed framework mandates that AI systems be built exclusively to serve human welfare and explicitly rejects the concept of machine personhood or model rights. Suleyman also called for greater transparency, independent scrutiny of AI behavior, and stronger tools to monitor and control technology, noting that autonomous multi-agent deployments have already exposed control vulnerabilities.