OpenAI Says It Expects That Future Models Will Be Able To Identify & Exploit Vulnerabilities Of Which It Is Not Yet Aware