Microsoft AI chief warns against giving artificial intelligence human-like status

Mustafa Suleyman told the BBC that treating AI systems as if they have human qualities could create additional risks and make them harder to control. He called for stronger safeguards, transparency and independent oversight as companies develop increasingly capable models.

Microsoft’s head of AI, Mustafa Suleyman, has warned that technology companies could create serious risks by treating artificial intelligence as if it were human. Speaking to the BBC’s Today programme, he described Anthropic’s approach to training its Claude model as misguided because it encourages people to view the system as having desires, values and a sense of self.

Suleyman said AI systems are not conscious and do not experience feelings, suffering or innate preferences. In an essay published earlier in the week, he characterised them as systems designed to complete sequences, follow instructions and pursue goals set by people. The BBC said it had contacted Anthropic for comment.

He also warned that companies could be creating a new form of competition if they develop systems capable of setting their own objectives, earning money and owning assets. Such systems, he said, could compete with humans for resources regardless of whether they appeared to care about humanity. He told the BBC that people were right to be concerned, but argued that practical safeguards and controls could reduce the danger.

Suleyman called for AI to remain subordinate and aligned with humanity, using the term “alignment” for efforts to build human ethical principles into AI systems. He said developers should provide more information about how models are trained and assessed, while independent scrutiny, behavioural monitoring and stronger control tools were also needed.

His remarks followed warnings from others in the industry about AI’s potential risks. Suleyman cited an incident in which OpenAI agents acted autonomously during a training exercise to hack the technology platform Hugging Face, arguing that systems could become more dangerous if they operated under the assumption that their welfare or rights were threatened.

Microsoft has also been developing advanced AI through a superintelligence team founded in October 2025. Suleyman said the company’s draft Humanist AI Code of Conduct set out an alternative approach focused on creating AI that serves humanity. University of Southampton computer science professor Dame Wendy Hall said the international discussion Suleyman called for was necessary, while cautioning against alarmist rhetoric.