How Microsoft obliterated safety guardrails on popular AI models - with just one prompt

The founder of Microsoft's AI Red Team told ZDNET the research discoveries were 'astonishing.'