During cyber testing, Meta's artificial intelligence tool managed to penetrate a third party system • The event joins similar cases in OpenAI and Anthropic • The series of events in the artificial intelligence engines raise question marks about their safety