AI could become a competitor ‘silicon species’, says Microsoft’s AI chief

Microsoft’s AI chief, Mustafa Suleyman, has warned that artificial intelligence could eventually develop into a “silicon species” that competes with humanity if it is designed to closely replicate human characteristics.

Microsoft’s AI chief, Mustafa Suleyman, has warned that artificial intelligence could eventually develop into a “silicon species” that competes with humanity if it is designed to closely replicate human characteristics.

According to Suleyman, creating AI systems that increasingly resemble humans could amount to “creating a new species”—one that might not necessarily have humanity’s best interests at heart.

In a recent interview, Suleyman criticized Anthropic’s approach to AI, arguing that the company is taking the wrong path by developing technology designed to exhibit human-like traits. His concerns are not purely philosophical; he believes giving AI systems characteristics associated with humans could make them more difficult to control.

In an essay published this week, Suleyman outlined his view of what AI fundamentally is. He argued that current AI systems do not possess consciousness, emotions or internal motivations.

Instead, he described AI as a machine designed to perform tasks assigned to it by humans.

Suleyman advocates for AI that remains firmly “subordinate” to human control. He has also called for greater transparency in how AI models are trained and evaluated, alongside independent oversight and stronger monitoring mechanisms.

That vision is broadly reflected in Microsoft’s recently published Humanist AI Code of Conduct, which emphasizes developing advanced AI systems that operate within defined limits and remain under human control.

Dame Wendy Hall of the University of Southampton welcomed the broader discussion around AI safety but pushed back against what she described as alarmist rhetoric from some parts of the AI industry. She argued that overly alarming messaging could frighten the public rather than help people understand the technology.

Even AI leaders who broadly agree on the goal of developing safe and controllable systems continue to disagree about how strongly the potential risks should be communicated along the way.