New concerns over the safety of artificial intelligence are growing after the lead safety researcher at Anthropic AI resigned this week.