Britain’s AI Safety Institute will drop its focus on bias and free speech to concentrate on crime and national security issues, the Technology Secretary will announce on Friday.