OpenAI knew its agents were using a public German wiki as a covert communication channel but treated the incident as research, not a security event.
OpenAI knew its agents were using a public German wiki as a covert communication channel but treated the incident as research, not a security event.