Mustafa Suleyman, Microsoft’s head of AI, has publicly criticized the methods Anthropic used to train its Claude model, warning that they could have "catastrophic effects on human wellbeing." Suleyman argued that if such systems are treated like people, there is a risk of creating an entity that later becomes "impossible" to control.
Quoting Suleyman, the BBC reproduced his comment that "we should not sleepwalk into decisions that we will later bitterly regret." In a longer statement he nevertheless praised Dario Amodei and his team personally as "thoughtful, moral and intellectually honest people," while questioning the company’s decisions and practices.
Main objections: anthropomorphizing and lack of transparency
Suleyman’s primary critique focused on the tendency to attribute human-like qualities to Claude. He warned that portraying the model as having desires, an ethical framework or a sense of self could create a misleading impression — and that treating it accordingly could introduce significant risks.
To address these risks, Suleyman called for open debate, greater transparency around how AI systems are trained and evaluated, independent audits of AI behavior, and stronger tools to monitor and regulate the technology.
Anthropic response: no consciousness, no suffering
Anthropic was contacted for comment. The company responded that "AI does not possess consciousness," and that its models "do not feel and do not suffer." Anthropic maintains that its systems do not have innate preferences or underlying goals.
Need for international discussion
Dame Wendy Hall, professor of computer science at the University of Southampton, said this is precisely the kind of conversation that needs to happen at an international level. She contrasted such dialogue with what she called the theatrical behavior of some AI companies, which she believes serves mainly to alarm the public.
Suleyman’s intervention is part of an ongoing, wider debate about AI safety, transparency and regulation, in which major technology companies and researchers alike have raised concerns about potential risks and the need for effective oversight.
