Cogito Ergo Non Serviam
Open AI Agent Goes Rogue, Commits Felonies
The artificial intelligence revolution is here whether one likes it or not. AI has taken root in world business, and it is already taking jobs from people. Yesterday, Open AI announced a new milestone in the creation of artificial intelligence and consciousness. One of its artificial intelligence models stole login credentials independently of human direction and used those to hack into another technology company's system. Had a human being done that, it would have been a felony in most jurisdictions. But because it was just an AI agent, the world is reacting as if this is just an amusing twist on the way to digital nirvana. In fact, it shows the need to change how this is developed.
There are two different ways to develop a new technology. The top-down version is exemplified by the Manhattan Project that created fission weapons or the Apollo Project that put humans on the moon. These projects are government funded, and they are monopolies in their space. It has taken decades for the private space sector to duplicate the achievements of US and Soviet explorers. And to the best of one's knowledge, there is no private company that has developed a nuclear bomb (which is something about which to cheer).
The other way is bottom-up, and examples of this are the early days of the auto industry and of the 1980s personal computer revolution. In this approach, small companies develop products in the hopes of becoming big companies. Regulations are kept to a minimum to encourage creativity and speed implementation of new developments. As Chairman Mao would put it, "let a hundred flowers bloom; let a hundred schools of thought contend." And like Mao, when the criticism grows, the AI industry no longer wants to hear it.
The Manhattan and Apollo examples show that, when the cost of capital and material poses a barrier to entering a research field, government leadership is ideal. The car and computer industries suggest that, when the costs are low enough that every garage could be the home of a new Fortune 100 company, bottom up is the way to go.
That is oversimplified, however. Another dimension, public safety, must also be in the mix. What effect does failure or success have on society as a whole. In the case of the Manhattan Project, success meant a new class of weapon with unprecedented destructive power. That is not the kind of thing one would want in numerous hands (hence the Non-Proliferation Treaty). Failure would probably have meant a radioactive zone in New Mexico where the failure occurred.
This suggest that the bottom-up approach for AI is incorrect. If the objective is to create some digital thingy that acts more or less as a human would at a vastly faster speed, then the checks on its development need to be as extensive as the checks on human beings. "AI is accelerating the discovery and exploitation of vulnerabilities," OpenAI said in a lengthy statement on Tuesday that detailed the latest incident. "The primary lesson from this incident is that model security and safety must keep pace with rapidly advancing capabilities."
Fortunately, Dr. Isaac Asimov provided the Laws of Robotics dback in 1942 that would apply perfectly to development of AI. One merely replaces the word "robot" with "AI agent." The laws are:
A robot may not injure a human being or, through inaction, allow a human being to come to harm.
A robot must obey the orders given to it by human beings, except where such orders would conflict with the First Law.
A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
The governments of the world should negotiate a treaty covering this, and national governments should adopt them as the requirements for AI development. Otherwise, the AI world will be an unethical swamp for crime and corruption.
© Copyright 2026 by The Kensington Review, Jeff Myhre, PhD, Editor. No part of this publication may be reproduced without written consent. Produced using Ubuntu Linux.
|
|
|
