Awesome stuff as always, Pavel! Thank you for the dope links.
Very curious about your deeper thoughts on this excerpt for your own use of agentic AI and everyone else's: "If that sounds awesome, then you’ll understand why I installed it! If that sounds like a security disaster, then you’ll understand why I balked at giving it any permissions. For better or for worse there are others that don’t seem to care, so we’ll see how this plays out!"
A caveat on my own slant. I very rarely use tools like ChatGPT a work and never do personally in part because I want to put in the real work when I can and in part because I just don't trust 'em from a security and privacy standpoint.
Taking that view on agentic AI, I am incredibly concerned about the security concerns you allude to. Signal's CEO Meredith Whittaker has great thoughts on these issues that are easily found on YouTube for the curious! I have seen 1Password start implementing certain security protocols around agentic tools but what concerns me is the "don't care to" line of yours above. I think even more people just don't know the risks so the question is how do we inform them properly while working to make these tools as secure and private as possible.
I've received enough breach notices in my life not to expect much privacy or security from anyone on the internet. I also expect car break-ins to happen >0% of the time, yet I still park my car on the street!
Awesome stuff as always, Pavel! Thank you for the dope links.
Very curious about your deeper thoughts on this excerpt for your own use of agentic AI and everyone else's: "If that sounds awesome, then you’ll understand why I installed it! If that sounds like a security disaster, then you’ll understand why I balked at giving it any permissions. For better or for worse there are others that don’t seem to care, so we’ll see how this plays out!"
A caveat on my own slant. I very rarely use tools like ChatGPT a work and never do personally in part because I want to put in the real work when I can and in part because I just don't trust 'em from a security and privacy standpoint.
Taking that view on agentic AI, I am incredibly concerned about the security concerns you allude to. Signal's CEO Meredith Whittaker has great thoughts on these issues that are easily found on YouTube for the curious! I have seen 1Password start implementing certain security protocols around agentic tools but what concerns me is the "don't care to" line of yours above. I think even more people just don't know the risks so the question is how do we inform them properly while working to make these tools as secure and private as possible.
I've received enough breach notices in my life not to expect much privacy or security from anyone on the internet. I also expect car break-ins to happen >0% of the time, yet I still park my car on the street!
For sure. For agentic AI though, the breaches would be across two or more systems and possibly financial. Not ideal!