Microsoft AI chief executive Mustafa Suleyman says AI safety requires more than aligning models to human values, arguing that containment and limits on AI agency are equally important. He warns that rapid progress could produce systems vastly more capable than today’s models, making regulation and safeguards increasingly urgent.

The discussion follows Microsoft’s 37-page “Humanist AI Code of Conduct” and Suleyman’s criticism of Anthropic’s views on AI consciousness and “model welfare”. He argues that AI proliferation is inevitable and often beneficial, but future systems could involve roughly 1,000 times more computing power than current models, creating greater risks of reward hacking, loss of control and harmful behaviour.