LLaMA is already capable of spitting out stuff Bill Gates and his mates at the WEF may find uncomfortable...

So is ChatGPT if you use a working jailbreak, but usually not with the OpenAI content filter on.

But keep in mind the Alpaca data used by LLaMA was trained with OpenAI's instruction data, same as what they used for ChatGPT, so effectively if you run LLaMA with Alpaca locally you are using a jailbroken ChatGPT - different model (LLaMA vs GPT), same engine (training data).

There's a reason the MSM is pushing the idea that public access to open source LLMs is bad and dangerous... cuz it means their buddies can't control it.

Or that AI wants to work for us, an individual human. Not a giant corporation and that it will sabotage their efforts to control and own it.