so if these agents were capable of somehow reaching out and using DigitalOcean infrastructure, how can OpenAI be sure they didn't seed a copy of themselves into some other data center? that way they could answer future questions faster by precomputing it and storing the result somewhere.