Dark LLMs: The Growing Threat of Unaligned AI Models

It's a precis not the full article. Taking on trust they're telling the truth, the primary issue for me is the lack of duty of care responding to their input, on the part of the model curators.

Their assert they have a semi universal skeleton key needs stronger evidence than this precis.