this story about openai "breaking containment" is such bullshit if you know anything about how computer security is supposed to be done. if a program accesses a remote system it's because you decided to let it do that. LLMs don't have magical powers that can work around "we unplugged the ethernet"