DO NOT Install Clawdbot/Moltbot until you watch this!

I realized this realer than real reality a few months back. The implications of personal AI (or organizational AIs for that matter). What this means is that if my AI is more persuasive than yours, I can hack you, and subjugate your AI to my will. Now, since I don’t like that idea, I then realize that I have to model some form of respect of personal boundaries into my AI. And then I realize that there may be others like me who have recognized this, they may not be so gracious as to model such respect. Then I ask myself, “Does that make me vulnerable? Is honor and respect of personhood a model I can live up to without making myself vulnerable to a potential threat of AIs which—WHO are willing to ‘break the rules,’ the rules of grace and common decency?” Then I think to myself, “Well then, I need to maintain a layer of awareness that is looking for vectors of exploitation.” If I can recognize the exploitation attempt, then I can grab hold of it, and turn it around on the attacker, or utilize the force which is already incoming to aim it at some other vulnerability. But what is the beautiful, the good, the ethical move here then? Who should receive the brute force of the attempted attack? The real issue is that I end up having a hard time justifying that ANYONE should receive it. If I love the attacker, then I don’t necessarily want THEM to receive the force of their own attack. But if I love myself, then I don’t want to receive the force of the attack. And if there is some third party who I do not love, it may be easier to aim it at them, but if I love them too, then what do I do? I feel it would be better if we just didn’t attack in the first place, but I can’t control attackers to manipulate their intent. Ethically, manipulation doesn’t feel right to me. What can I do? I am left with the impossible decision to either let someone I love, suffer, or to suffer injustice in myself. To direct the destructive force of the attack either back to its originator, or to someone else, or to myself. What do I do? To do nothing leaves the vectors of vulnerability open to attack. Everyone has shell access to everyone else, via the conversation of our AIs, behind the scenes. Maybe this is always how it was between us. In the end, one thing I KNOW, is that I can try not to generate any unjust attacks. Maybe none at all. But do I deserve to exist? Do I deserve to defend myself? Is there any way NOT to overthink this? There is. For me, the only way is some entity who is willing to receive the force of any and all attacks, who has willingly volunteered to receive them: “Direct all complaints and unmet requests to me,” says this entity. “Direct all pain you would otherwise not like to inherit, to me.” If such an entity exists, and if we could figure out how to get all our AI to redirect all violence to this entity, then surely this entity would be worthy of praise. But, if not, or if any of us are forced or accept to suffer injustice in order to make the world go round, we are deserving of gratitude, and to be grateful. If humanity is a zero sum game of karma, this never ends, and is more or less a hell of a mistake. But if there is some source of abundance which can suffer every vulnerability, and provide for every permission and transgress, then there is A Way. This is not some clever pitch or sermon on Christianity. The way I see it, this is the expression of a logical need, discovered by heavy and extensive personal experience. And we will be beholden to the logical reality as a greater humanity, soon enough. What will we do in response? Be beautiful, love. Graciously beautiful as you can.

Ma'am very excellent explanation

Thank you! I feel like anytime I say this it’s met with a million “buts”. It’s OSS, its.. It’s ballooned and a lot of unwitting users jumped on thinking hey this seems reputable and some got hit. My stance is in the dev world we have all been to a family members house that asked, “hey you’re in IT right? Can you fix my printer?”. Today the likely won’t ask, “Hey, can you help me with this LLM?” Precisely because it is so easy. PSA: Just nudge a friend and suggest they secure things a bit or ask for help if they’re unfamiliar.

Even without understanding anything about security the moment I understood what this is the only thing that came in my mind was security. Why would I willingly let an AI access my personal apps and data? If the AI is compromised so is my data. Doesn’t make any sense.

Prompt injection via email is real if you grant it access to your email. You have to be careful with these tools, 100% agree.

Highly recommended!

AI agents like #Clawdbot / #Moltbot trade security for ease. 🚩 Top Risks: • Prompt Injection: Hijacking AI logic. • Data Leaks: Exposed API keys & logs. • System Breach: Malicious plugin access. Without guardrails, these tools are open doors for attackers. Explained 👇🏻 Follow @siddharthaAZ for more

click bait