← Back to context

Comment by rf15

2 days ago

> As grounded as this article comes across

It's a post from OpenAI, so it is an advertisement piece.

How would you want them to behave? Suppress the news?

  • Well making an actual sandbox before testing offensive abilities of their supposedly _really dangerous model_ would have been nice...

    • They probably did, and it probably went along these lines:

      > Very dangerous model, please implement an unescapable sandbox.

      > Certainly, here's an unescapable sandbox!

        function execInUnescapableSandbox(cmd) {
          if (cmd.split()[0].startsWith("cd") && !cmd.split[1].startsWith("/sandbox"))
             throw new Error("[usbox] Rejected!");
          exec(cmd);
        }