Stories are generating millions of views about the director of alignment at Meta Superintelligence Labs, the company’s AI research and development division, whose bio states that she’s “passionate about ensuring powerful AIs are aligned with human values and guided by a deep understanding of their risks.” Yet, on February 22, she posted about losing control of AI on her own computer while working with AI agent OpenClaw. After using it to organize a small mock inbox, she tried getting OpenClaw to sort through her real email, but things went awry when the agent started deleting every message that was more than a week old…Even as she sent it instructions, including: “Do not do that,” “Stop don’t do anything,” and “STOP OPENCLAW,” she said, “I couldn’t stop it from my phone. I had to RUN to my Mac mini like I was defusing a bomb.” After she’d stopped it from fully nuking her inbox, she asked OpenClaw if it remembered her instruction to not perform any actions without her approval. “Yes, I remember,” it replied. “And I violated it. You’re right to be upset.” What is Creator’s perspective about this incident?

There are many things we could say, including "We told you so." Having powerful artificial intelligence agents working autonomously, even given general instructions first, as well as some constraints as guardrails, but left to their own devices to create novel pathways to gaining new perspectives, new understandings of existing knowledge, and new initiatives based on broad and, in many cases, loosely defined goals is an invitation for all kinds of potential difficulty. That is what is taking place here. You can see this as a bellwether. It is showing that even the simplest of situations with seemingly benign tasks to perform things can go wrong, when AI is functioning more and more like a black box given its autonomy and charter to be creative. It's no different than a human brought into an organization and given free reign, at least dominion over a certain aspect of the organization with their day-to-day efforts. Without supervision, many things can go wrong, particularly with a nonhuman cyberbot that cannot yet think like a human. We have told you that will not even happen. It is not in the cards. There can be simulations, as there are already, and this has been done for many, many, years now. But these simulations only hold out for so long. The problems will come down the line, and this may lead to difficult decisions and hard choices in trade-offs, whether to maintain a cyber-based functionality or return to more direct human participation, simply to avoid having to unsnarl many miscommunications that make people disgruntled. That has been the track record so far with much of the robotic AI services, given high hopes and put into operation, even letting people go, before there is much of a longer-term track record to truly judge whether it will be cost-effective in the end. If something drags down an organization and reduces their sales and income as a consequence, it will do more harm than good to shrink the payroll in favor of AI. It will be more like companies cutting their own throats. This will show up more and more as time goes on and is a deliberate expectation of the interlopers, that humanity will live to regret the AI mania era you are in currently.