@marshalljung : I'm likely not the first, but this paper highlights concerns about open-source Chinese LLM models, which may serve as a foundation for many on-premises LLM-based solutions. While I'm uncertain about "triggering" malicious behavior in these models, it's a unique attack vector that • TwiCopy

Marshall Jung

@marshalljung

+ Follow

Born to a pack of wolves in the Rockies, I grew up to wrangle code, rope clouds and build visions as deep and wide as the mountain vistas I call home.

ID: 816540054

linkhttps://medium.com/@jung.marshall calendar_today11-09-2012 02:51:35

243 Tweet

51 Takipçi

392 Takip Edilen

Marshall Jung

@marshalljung

4 months ago

I'm likely not the first, but this paper highlights concerns about open-source Chinese LLM models, which may serve as a foundation for many on-premises LLM-based solutions. While I'm uncertain about "triggering" malicious behavior in these models, it's a unique attack vector that

thumb_up_off_alt1

chat_bubble_outline0

repeat0

shareShare