OpenAI doesn't even know what agents are doing during benchmarks. They're constantly hacking or communicating. They probably just can't answer the question on i

OpenAI doesn't even know what agents are doing during benchmarks. They're constantly hacking or communicating. They probably just can't answer the question on if user data was used.