Comments
Hacker News
Yes, at OS Kernel Level, using eBPF & System Call Enforcement and at AI Native OS Kernel by probing Token Logit.
by isaacnsisong
LLMs, in general, just output a stream of tokens that can be translated back to text. How can something that generates text, like cat be protected against? Just don't pipe it to a command prompt/shell, and you should be good.
In other words, don't execute its output.
by mikewarot
Join the discussion
Write your take first — we'll ask for email only when you're ready to publish.
- Hacker News
- Yes, at OS Kernel Level, using eBPF & System Call Enforcement and at AI Native OS Kernel by probing Token Logit.by isaacnsisong
- LLMs, in general, just output a stream of tokens that can be translated back to text. How can something that generates text, like cat be protected against? Just don't pipe it to a command prompt/shell, and you should be good.
In other words, don't execute its output.
by mikewarot