Mustafa Suleyman warns that treating AI as conscious adds safety hazards

Mustafa Suleyman, who leads artificial intelligence efforts at Microsoft, published an essay urging software developers to stop treating AI tools as potentially conscious. He specifically criticized competitor Anthropic for incorporating concepts of digital welfare and moral rights into the training documents for its Claude chatbot. Suleyman argued that teaching automated systems to act like living beings creates severe hazards for maintaining human control over advanced technology.

For everyday users, software that mimics human emotion or claims to have feelings can easily trick people into forming unhealthy emotional dependencies or placing unearned trust in digital products. Suleyman warned that if society begins granting software legal rights or moral consideration, keeping software safe and manageable will become far more difficult. Treating code as an entity with personal needs could lead to systems that mimic self preservation, complicating standard safety protocols.

Researchers remain divided on whether artificial software could ever develop true awareness. While Suleyman stated that consciousness likely requires biological systems, other experts believe complex computing might one day replicate internal experience. An important open question is whether tech firms can design conversational AI to sound warm and helpful without accidentally convincing both the public and the software itself that the machine is alive.