AI needs "rigorous" testing and safeguards to contain risk, Andrew Bailey says.

This story was reported by BBC News. Read the full coverage at BBC News.