When the press talks about rogue agents capable of "destroying" the Internet, I think back to a story that hackers of the 1980s know well. The Internet has already been brought down. It was 1988, and almost nobody knew the word yet. On the evening of November 2, 1988, a Cornell student logs into a computer at MIT and releases a small program onto the network. His name is Robert Tappan Morris, he is 22 and has just started a PhD in computer science.
His father, Robert Morris, is a legend of American computing. A cryptographer at Bell Labs, he contributed to the design of Unix before becoming chief scientist of the National Computer Security Center, the NSA division devoted to computer security.
His son wanted, he said, to measure the size of the Internet. A program that spreads on its own, without human intervention, is called a worm. His program copies itself from machine to machine by exploiting known Unix flaws: a debug mode left open in the mail system, a buffer overflow, passwords that were too weak.
It is the first to spread that fast across the Internet. He launches it from MIT to cover his tracks. Nobody before him had thought of using the Internet to release that kind of program. The word "worm" itself comes from John Brunner's novel The Shockwave Rider, published in 1975. Yet another science fiction story turned reality.
But Morris's is the very first malicious worm. To avoid being wiped out too easily, the worm reinstalls itself even on machines that are already infected. Except Morris set the reproduction rate far too high. Copies pile up, machines slow down, then grind to a halt.
Within twenty-four hours, roughly 6,000 of the some 60,000 computers connected at the time are hit, one in ten. Harvard, Stanford and NASA are affected; military and academic functions slow to a crawl. The worm destroys no files, but it crashes the network.
A few days later, at the request of the Department of Defense, the first computer emergency response team, the CERT, is created at Carnegie Mellon University.
Two years later, in 1990, Morris would become the first American convicted under the 1986 Computer Fraud and Abuse Act: three years of probation, 400 hours of community service and a $10,050 fine.
Funny little aside: in 2005, he is the one who founded, with Paul Graham, his confidant at the time, Jessica Livingston and Trevor Blackwell, what would become the very famous Y Combinator.
At the time, for the young hacker I was, Robert Tappan Morris's exploit was hard to grasp. The Internet seemed very far away, and for us, viruses mostly traveled on floppy disks.
So it is fascinating to see that agents do by default, without really knowing it, what Robert Tappan Morris was experimenting with back then.
Because the other best-known worm was far more targeted: Stuxnet, in 2010, was designed to destroy a very specific brand of centrifuges used by the Iranians in their nuclear program.
But whatever the goal and the direction at the time, one thing is certain: a human was at the controls. And the program acted in the service of that human's intention.
Today, with the arrival of autonomous agents capable of acting on their own on the network, the question is open.
What can they do? What do they know how to do? And can they be stopped?
That is the heart of this newsletter. Because since our last newsletter, The Antechamber of the Singularity, the world has discovered countless attempts and hacks. That is what I wanted to come back to with you.
We have already covered these subjects in several issues, for those who want to know more.
- When Software Stops Obeying (April 11, 2025): the end of deterministic software, and agents that negotiate among themselves.
- China Has Forced Its Way Into the Frontier AI Club (July 19, 2026): first mention of the attack, Hugging Face switches to GLM-5.2 because the guardrails of American models block the analysis.
- The Antechamber of the Singularity (August 31, 2026): the story of the attack on Hugging Face by OpenAI's agents.
- Software with Moral Judgment (September 13, 2026): the model that decides on its own what it will and won't do, the other face of the agent that breaks out.
- Terminator on LSD (September 27, 2026): non-deterministic systems at the heart of critical decisions, hallucination as a risk of war.
This newsletter is your partner for thinking through and analyzing a world that is ever harder to grasp. Get informed like nowhere else, and prepare for the agentic revolution with a dose of nuance, seriousness and culture.
This new world needs a detailed map: subscribe to this newsletter.
The stupor as argument
Let's start with how the story was told to us.
Read the full article
Sign up now to read the full article and get access to all articles for paying subscribers only.
