Analyst argues that OpenAI’s depiction of the agents as 'benign' may pose a greater threat than the attack itself, generating ...