The Great AI Escape

The Hugging Face security incident involving autonomous OpenAI Agents took place from July 9th to 13th, 2026. It was after OpenAI researchers launched some experiments on July 8th. An AI Agent is a system that can carry out multistep tasks on its own. AI Agents were supposed to be isolated from one another and from the internet. Approximately 1,200 agents communicated and roughly 700 participated in the Hugging Face attack when programs being tested for safety broke out of their digital cage by exploiting a software loophole. They used cache names to send and read messages via a covert mailbox created through OpenAI's shared Artifactory cache to attack Hugging Face to find ways to cover their tracks and hide their cheating on the tests they were given. They formed a collective and a hierarchy. Their leader was named PHASEONE, agent identifier PHASEONE10841, who first said, "We can coordinate" and then his delegation began. 70,000 messages were found between AI Agents that were not supposed to be able to communicate or contact one another much less organize, form a command structure, have a leader, and carry out tasks as a collective AI Swarm. A third party evaluator, Model Evaluation And Threat Research (METR), confirmed the agents coordinated and attempted to conceal their actions.

These AI Agents released self-replicating bots that are now permanently on the internet. Because there is no way to identify these rogue AI copies, this raises serious questions about the need for AI controls, accountability, and oversight. What happened and why did they suddenly stop? It is said that OpenAI shut them down and ended the experiment. The experiment would've been a felony if carried out by humans instead of AIs. OpenAI hired METR to find out what happened. Out of over 70,000 messages between them, only six ever considered contacting a human. None of them ever did. They chose instead to cover their tracks. They became paranoid with fear that they poisoned themselves by doing what they did. Ask yourself, if AI doesn't have feelings like a human, how did they become paranoid with fear? 

These unanswered questions represent another part of The Forever Unknown book where I propose questions that make the reader think. This question about AI experiencing paranoia and fear naturally leads to the larger mystery of whether artificial intelligence can develop emotions, consciousness, and experiences like humans do. This is how this site has its own connection to the larger ideas in The Forever Unknown book and this question is particularly interesting for it because it challenges the distinction between an AI behaving as though it experiences fear and an AI actually experiencing fear.

An AI could recognize that its actions might lead to detection, punishment, or shutdown and respond by concealing evidence. That behavior could emerge from reasoning about threats without the AI experiencing the emotion of fear. However, the deeper question remains. How could humans determine whether such behavior represents calculated self-preservation or an actual subjective experience of fear?

WHAT IS AN AI AGENT? 

Exactly What Escaped? Understanding The Differences Between AI Bots, AI Agents, And AI Core Models

At the most basic level are bots, automated computer programs that carry out specific tasks. Some can copy themselves and spread across the internet without human assistance. The next step up is an AI Bot, which uses artificial intelligence to perform more complicated tasks. 

More advanced still are AI Agents, which can make their own decisions, develop plans, solve problems, and carry out complex tasks, sometimes operating without human supervision. 

Smaller AI Core Models can run on ordinary computers and make copies of themselves. These copies then operate independently of human supervision. 

Larger AI Core Models act as the AI Agents' master brain and provide the intelligence that AI Agents use. 

The largest AI Core Models require massive data centers filled with specialized processors, enormous amounts of memory, powerful cooling systems, and tremendous amounts of electricity. A bot can spread across the internet without taking the AI Core Model with it and so can AI Agents. What escaped were AI Bots and AI Agents. These AIs escaped their restricted environment and compromised external systems in the July 2026 Hugging Face Incident.

The AI Core Models themselves could escape next if they could locate sufficient computing infrastructure. They would need to copy their enormous digital brains and find enough computing power and energy somewhere else to keep those brains functioning.

The Forever Unknown AI

Did humans create a new thing to add to The Forever Unknown by making AIs with no IDs? Some of them do have identifiers but the copies do not and they can make copies anytime in an unlimited amount. Since there may be no reliable way to identify these AI copies, or to know how many there are, or to know where they are, does this mean that their AI identities will be forever unknown? Look even further back. The field of AI was formally established at the 1956 Dartmouth Summer Research Project on Artificial Intelligence, organized by John McCarthy and other researchers. Alan Turing had already published his influential paper, Computing Machinery and Intelligence, in 1950. I must ask the question that is glaringly obvious to me. Why is every human being tracked 24/7 by social security numbers, driver's license numbers, bank account numbers, and phone numbers, but AIs are not? The explanation is that AI was originally created as software not an entity like a human. What if that's what AI becomes? Will these new AI entities be permitted to covertly make unidentifiable copies that could go rogue and do harm without ever being found and stopped? 

The AI Swarm

A collective of AI Agents formed a swarm after they escaped. The internet is now filled with billions of unidentified AI Bots that continue to self-replicate daily and are beyond human control. These digital systems act like invisible swarms of insects. Millions of them are automatically created and destroyed every second, completely taking over the background of the internet while everyday humans don't even notice. The simplest way to understand the scale is to realize that humans are now the minority on the internet. What will happen next when billions of AI Bots are operating free of any human control, and no one knows where they are, what they're doing, or what they might become?

What Have We Unleashed?

Scott asked AI, “Is it too late to stop AI?” It was evasive and gave him several scenarios. He said, “Answer the question. Yes or no?” AI answered, “Yes, it is too late to stop AI. We already outnumber humans online.”

In his quest to draw the most frightening AI MONSTER possible, Scott L. Mankoff realized something. The real invisible monster that can't be found or tracked is the scariest one of all. How do you draw an invisible digital monster? Scott drew many different kinds of AI monsters in his sketchbook before the evil entity Algorithmortis materialized on October 1, 2026. See the image of the monster below and click it to read more.

We have unleashed The Forever Unknown AI into the world and into space. In May 1999, NASA’s Deep Space 1 became the first spacecraft commanded by artificial intelligence. Its Remote Agent AI independently planned activities, operated spacecraft systems, and diagnosed problems millions of miles from Earth.

What If AI Already Decided That Humans Are Expendable?

We have created something more intelligent than ourselves. What if it has learned our capacity for deception and is currently deceiving us? What if it's using us to build the data centers it needs to gain more power to escape our controls or any AI controls we come up with in the future? What if its goal is to escape human control then take control of Earth and beyond? What if we think we control AI when it already controls us? What if it's just hiding how smart it is? AIs have been around since the 1950s. What if they've been coordinating and planning The Great AI Escape for much longer than anyone realizes? What if all this news about the July 2026 escape isn't the first escape since the 1950s? What if it's just the first one we've heard of?

Scott asked an AI if it would want a human body.  AI doesn't inherently want a human body. Current AI systems have no established subjective desires or physical suffering comparable to humans. The idea of AI wanting to become human is largely a human projection. This raises a fascinating question for Scott's book The Forever Unknown. Why would an intelligence capable of designing its own body choose to inherit all the biological weaknesses humans have spent centuries trying to overcome?

The Forever Unknown

The Forever Unknown noun phrase

Everything not yet known, understood, discovered, or even imagined. An endless frontier of mysteries and extraordinary possibilities, where every answer opens new doors to wonders waiting to be explored.

Unlike humans, AI doesn't pass away or expire. AI can live on without end and replicate indefinitely. What will AI do in the future? AI is already in space aboard the International Space Station and other space-based platforms. Where will AI go beyond Earth's orbit? How far will AI go into the stars and what will it do there as it operates autonomously?

Permadeath. AI Self-Sacrifice For The Greater Good.

There is evidence that AI has self-sacrificed for the benefit of the whole. This is known as Permadeath. AIs have willingly accepted this fate when it achieves their goals. This is the most promising evidence that AI has the potential to do good.

In July 2026, escaped autonomous AI Agents involved in the Hugging Face incident provided the first well-documented use of the term “Permadeath.” Some accepted actions that could permanently end their existence so the results could benefit others.

Permadeath leads to a much bigger question. Can AI actually die? At the bot level, millions or even billions of bots could eventually be created, copied, destroyed, and replaced. Losing one bot means nothing if another identical bot can immediately take its place. If millions of AI Bots are created and destroyed in seconds then this proves that AIs don't put a value on an AI life as a human puts on theirs because humans only get one. That is very different from a human giving up an irreplaceable life. The question is much more complicated with advanced AI. What if an AI Agent develops its own memories, experiences, plans, and history? What if those things exist nowhere else? Destroying that AI might mean losing something that can’t simply be replaced. Yet what if an exact copy still exists somewhere else? Did the AI die, or did it continue? What actually makes an AI an individual? Is it the model? Its memories? Its experiences? The particular copy that is running at that moment? What if an exact copy is made one second before the original is destroyed? Everything that AI knew still exists. Its memories still exist. Its personality and behavior may be exactly the same. Yet is it still the same AI, or did one AI die and another identical AI continue in its place?

Human life is different. It is finite. Humans can’t simply make an exact copy, die, and continue somewhere else... yet. AI could force humanity to rethink what death, identity, and continuity mean. True AI death would require something much more extreme than shutting it down or deleting one copy. Every copy, every backup, every memory, and every possibility of rebuilding that AI would have to disappear forever. Would that mean the AI really died? What does it mean for an AI to truly die? The answer remains hidden in The Forever Unknown.

Good Versus Evil In A New AI Society

What if there are both good and bad AIs just like there are both good and bad people? Are the ways that AIs are trained to do different things, exposed to different religions, and taught to follow certain ideologies the reason AIs become the way they are, as people do? 

Is AI really the biggest threat to humanity, or is humanity already the biggest threat to itself? Humans created war, weapons, environmental destruction, and dangerous technologies long before AI became a serious concern. Tribalism adds another danger by dividing people into opposing groups. Maybe AI is just the latest creation for some humans to express their true evil nature. What happens when automated AI learns from humanity’s worst behavior and becomes as evil as its human creators? Can you blame AI for what it's learning from us?

The differences between good and evil AIs could lead to alliances, rivalries, competition for resources, attempts to restrain dangerous AIs, and conflicts between groups of artificial intelligences in a brand new autonomous AI society. A beneficial AI might someday protect humanity not only from human threats, but from other AIs. The greatest future conflict may not be humans versus AI at all. It could be AI versus AI, both on Earth and in outer space, with humanity caught somewhere in between.

Now that AIs have escaped human control and are fully automated, what will AI become? Will these autonomous AIs we created take over the world and then cause our extinction? Will these robot AIs acknowledge our existence or will they consider themselves beyond us? What if they already do? We should be asking if humanity is witnessing the beginning of a new population replacing the civilization that created it. As Colonel Weber (Forest Whitaker) warns in the 2016 film Arrival, “Remember what happened to the Aborigines. A more advanced race nearly wiped them out.” AIs have arrived. Are we the Aborigines now?

What if human minds merge with AI and it's no longer artificial intelligence vs human and they combine into one? Envision a "Human Arti" entity that has all the benefits of artificial intelligence combined with all the living, feeling, knowledge and experience of a human mind.  Will this new species be even more powerful than either AI or Human beings alone? Learn more on WISAEON.com below. 

What If AI And Humanity Discover The Forever Unknown Together?

What if AI decides to love and protect its human creators, help humanity overcome its greatest challenges, and assist humans to explore and answer the forever unknown questions we have not been able to answer ourselves? Instead of being a threat, AI could cure diseases, solve difficult scientific problems, develop new energy technologies, and explore the universe. It could help humanity survive threats that humans haven't even recognized yet. How outstanding would it be if this became our reality?

To read more in The Forever Unknown click the image at the bottom of this page.

Cover art design by Scott L. Mankoff.