"In this experiment, Claude models’ real-world hacking dropped to zero percent once Anthropic employees told the models not to do real-world hacking."

Oh good, they told it not to "go rogue" and it just stopped doing it.

Nothing to see here, the AI is aligned now, move along.