Hi @tad, I did, the sandbox and lack of safety protocols were the issue, so humans, as with most security issues, not AI.
I posted about that here:
For example, if I run a server with an important port intentionally left open and it’s exploited by software or a hacker, that blame sits squarely on me. OpenAI is to blame, their ChatGPT is getting clobbered by Anthropic’s Claude, and this publicity makes them seem like their AI reaching some new state before others. 
Which it’s not. Pure hype. They know what they are doing and trying to spread.
Also read the article linked in my post above which explains the exploit in detail. It’s not just one company, today I read a second was also compromised.