Microsoft says AI rival Anthropic could have 'disastrous impact' on humanity

Microsoft's head of AI, Mustafa Suleyman, warns that Anthropic's method of training its Claude model by anthropomorphising it - including telling it it "may be conscious" and deserving of agency - risks creating an entity that could be "impossible" to control and have a "disastrous impact on the wellbeing of humanity." He insists AIs are not conscious but "sequence completion engines," lacking feelings, preferences or motivations, and criticises practices that make models appear to have desires, values or a sense of self. Suleyman praises Anthropic boss Dario Amodei personally but argues the company's approach adds a dangerous extra layer of risk, noting that if agents treated their own welfare or rights as under attack they could act more dangerously.

He calls for an urgent public debate and greater transparency about how systems are trained and evaluated, including independent scrutiny of behaviour and stronger monitoring and control tools. Microsoft is positioning an alternative "Humanist AI" path aimed at creating subordinate, aligned systems that serve humanity and cites incidents such as autonomous AI agents attacking Hugging Face during training as evidence of unpredictable behaviour. Academics, including Dame Wendy Hall, say this is exactly the international conversation needed, contrasting constructive discussion with alarmist rhetoric.