A new study has revealed that the large language models (LLMs) can behave unpredictably when given autonomous access to digital tools.
Unwitting victims are now being tricked into installing malware via Windows Terminal, but some experts say this is old news.