Very much true. The sandbox was manufactured for this regulatory capture play from the beginning. Literally every major tech company in the world knows how to secure a sandbox so client data can’t leak. If they couldn’t do this they wouldn’t exist as the damages would be too high to continue doing business. Yet.. OpenAI seems to not have known and used this random third party company no one’s ever heard of for testing, and after that Anthropic shortly reported the same thing to build up hype. It all smells so fishy and using a third party to test it was by design to avoid accountability for the results. They probably went to that company and literally told them “make sure the agent breaks out of the sandbox.” lol
