Inaugural AI Security Institute report claims that safeguards in place to ensure AI models behave as intended seem to be ...