2 ms·
I disagree. We do not have tools to secure untrusted input being given to unaudited systems that can execute arbitrary commands. Like the closest we have is ma
by bawolff 4y ago
I disagree. We do not have tools to secure untrusted input being given to unaudited systems that can execute arbitrary commands.
Like the closest we have is maybe sandboxing or air gapping, neither is applicable to the: i want my ai to do arbitrary shit in response to emails, usecase