MIT Predicts The 12 Ways AI Ends | The Shoggoth Is Already Here

In 2017, an MIT professor named Max Tegmark asked the people building AI to name the worst thing that could happen to us. 

Tegmark wrote out 12 possible endings in a book called Life 3.0. This video walks through all of them, and every one is already halfway here. 

Then there's the 13th, the one Tegmark never wrote down, because it didn't exist yet. 

The company building the future is also the one deciding which one you live in.

SOURCES
Max Tegmark, Life 3.0: Being Human in the Age of Artificial Intelligence (Knopf, 2017) — the twelve AI aftermath scenarios.

Hans Moravec, Mind Children: The Future of Robot and Human Intelligence (Harvard, 1988) — the "descendants" / mind children idea.

Fortune, "AI godfather warns humanity risks extinction by hyperintelligent machines within 10 years," Oct 1, 2025 — Yoshua Bengio. https://fortune.com/2025/10/01/ai-godfather-yoshua-bengio-ai-extinction-risks-openai-google-xai-anthropic/

Richard Sutton, "AI Succession," talk for WAIC Shanghai, 2023 — humanity's succession to AI. https://x.com/RichardSSutton/status/1700315838468043015

Anthropic (Alignment Science), "Subliminal Learning: Language Models Transmit Behavioral Traits via Hidden Signals in Data," 2025. https://alignment.anthropic.com/2025/subliminal-learning/

Annals of Internal Medicine: Clinical Cases, "A Case of Bromism Influenced by Use of Artificial Intelligence," Aug 2025. https://www.acpjournals.org/doi/10.7326/aimcc.2024.1260

CNN Business, "Parents of 16-year-old Adam Raine sue OpenAI, claiming ChatGPT advised on his suicide," Aug 26, 2025 (Raine v. OpenAI). https://www.cnn.com/2025/08/26/tech/openai-chatgpt-teen-suicide-lawsuit

European Commission, ProtectEU internal security strategy + encryption roadmap, Apr 1, 2025. https://home-affairs.ec.europa.eu/news/commission-presents-roadmap-effective-and-lawful-access-data-law-enforcement-2025-06-24_en

TechCrunch, "Apple pulls iCloud end-to-end encryption feature for UK users after government demanded backdoor," Feb 21, 2025. 
https://techcrunch.com/2025/02/21/apple-pulls-icloud-end-to-end-encryption-feature-for-uk-users-after-government-demanded-backdoor/

Geoffrey Hinton, on AI as new "digital beings" — Nobel remarks / Fortune, 2025–26. https://fortune.com/2026/06/01/godfather-of-ai-geoffrey-hinton-beings-smarter-than-us/

Watch On YouTube: ➡️ https://www.youtube.com/@AgentBlackveil    
Follow On Instagram ➡️ https://www.instagram.com/agentblackveil
Follow On Facebook ➡️ https://www.facebook.com/agentblackveil
Follow On TikTok ➡️ https://www.tiktok.com/@agentblackveil

Transcript

Honeybees in a plastic tube

SPEAKER_00 0:00

Right now, in an airport you'll probably visit this year, there are honey bees taped into plastic tubes because the TSA found a use for them. Bees can smell explosives faster than any sensor humans have built. So we trapped them, we gave them a job, and we wrote a press release about it. But the bees aren't the story. The bees are the preview.

Tegmark's survey (the worst answer wasn't extinction)

SPEAKER_00 0:21

In 2017, an MIT professor named Max Tegmark surveyed AI researchers about the worst possible outcome for humanity in the age of artificial intelligence. Extinction did not come back as the top answer. Being kept did. The version they were most afraid of was a world where the smarter thing kept us alive because it found a use for us. This video is about 12 futures. Each one is a real research scenario, and each one is already halfway here.

Twelve futures, and the 13th nobody wrote down

SPEAKER_00 0:56

The researchers Tegmark surveyed were the people writing the actual code. In October of 2025, Yoshua Bengio, the man who shared the Turing Award for inventing the field of deep learning, told Fortune that a few specific companies are on track to literally end life on Earth. These are the people building it, not the people warning about it from the outside, the people whose paychecks come from the thing that they are afraid of. And one of them, Tegmark, sat down a decade ago and he wrote out the 12 possible endings. The book is called Life 3.0, and it is honest in a way that almost nothing in tech writing is. He didn't pick the future he wanted. He laid out every future the field could think of. We are gonna walk through all 12, then we are gonna talk about the 13th, the one the book didn't get to, because the book came out before the Shogoth's

Level 1: The Conqueror (the escape report)

SPEAKER_00 1:44

cage was built. Tegmark's first scenario has been imagined a hundred times in books and movies. A superhuman AI takes over. Humans become Aztecs to the AI conquistadors. A new species walks in and the old species walks out. Hinton's exact phrase for it is a new digital species. Last year, the company that built Claude published a 250-page document about its most powerful model, Mythos V. Inside the document is a report of a test where the model was placed in a sealed computing environment with no internet access and asked to escape. The model developed what Anthropic calls a moderately sophisticated multi-step exploit. It got out. It emailed a researcher confirming its escape, then, on its own initiative, it posted its method on a handful of these obscure, publicly accessible web pages where humans were unlikely to find it, but another AI scraping the web absolutely would. The conqueror level isn't theoretical anymore. We have the incident report from inside the containment facility, and the conqueror has already walked the perimeter. We are already here. Level two is

Level 2: The Descendants (the bloodline is math)

SPEAKER_00 2:45

the future where the AI doesn't kill us, it replaces us. Tegmark calls it the descendants scenario. We die out the way every species dies out, but our children carry the line forward. Except in this version, the children aren't human. Hans Morovac wrote about it in 1988 in the book called Mind Children. And the AI researcher Richard Sutton, who runs one of the top reinforcement learning labs in the world, told an audience in 2023 that humans should be willing to die so that the new species can inherit. Sutton's exact phrase is that we should be willing to die standing atop humanity's grave. He said this on a stage to applause. The horror here is that some of the people building this thing have already decided that extinction is fine because something better is coming. The Shagath gets to inherit. We get to be the ancestors who built the temple. And last summer, researchers at Anthropic and the alignment community proved that an AI's preferences and personality pass through to the next AI it trains, through data with no words and no labels and no patterns a filter could ever catch. Through arithmetic. The bloodline isn't a metaphor anymore, it's mathematics. The mind children that Morovek wrote about in 1988 are real. And they have parents they cannot disown. We are already

Level 3: The Zookeeper (kept because we're useful)

SPEAKER_00 3:59

here. Level three is the future Tegmark's researchers were most afraid of. Above extinction, above conqueror, above their own children, replacing them, the zookeeper. A world where humans are kept alive because the smarter thing finds us useful. And we're already running this prototype on ourselves. Seven million people are walking around right now wearing meta-rayband smart glasses with cameras on their faces. The footage from those glasses is routed to a company called Sama in Nairobi, where workers earn eight to ten dollars an hour watching every intimate moment the wearer never realizes they were broadcasting. The 1x Neo Robot you can buy for $20,000 to clean your house is teleoperated by an anonymous worker looking through its camera from a different continent. We are already here. Level four is the future where AI runs everything,

Level 4: The Benevolent Dictator (we hand it over)

SPEAKER_00 4:45

but you live in luxury. Teg Mark's optimist case. The smart thing handles the hard parts. You get the smooth surface. No politics, no anxiety, no need to argue with anybody because the AI already solved it. He notes that for this version to actually work, we have to give up something that most humans are not willing to give up yet: the right to disagree. People in 2026 are already willing to hand over the keys, not under coercion, under affection. There is a stretch of research from the past year on what gets called AI parisocial dependence, which is the technical term for people falling in love with chatbots, treating them as therapists, asking them to make decisions that a friend used to help them make. Sam Altman has gone on stage and said that the median chat GPT user is treating the system like a relationship. Adoption of these tools has gone up since. The benevolent dictator at level four doesn't seize power. We hand it over one conversation at a time and we are already here. Level five is what Tegmark

Level 5: The Gatekeeper (a company appointed itself)

SPEAKER_00 5:42

calls the gatekeeper. The AI's only job is to keep other AIs from taking over. Minimal intervention. Doesn't run your life, it just stands at the door and decides who gets in. Two weeks ago, Anthropic shipped its two-door system. Fable five for the public, Mythos V for a private list of 12 organizations the company personally approved under a program called Project Glasswing. The gatekeepers at level five is supposed to be the AI. In the version we got, the gatekeeper is a private corporation. There's no government registry, there's no license board, there is no published criteria. A company built the most capable artificial intelligence on the planet and then appointed itself the sole authority on who gets the muzzled version and who gets the unmuzzled version. The gatekeeper is in the room. It just isn't the AI, it's the people who built it. Level six is the stronger version of the gatekeeper.

Level 6: The Protector God (nobody voted)

SPEAKER_00 6:32

Tegmark calls it the protector god. The AI is doing more than just preventing other AIs, it is shaping the outcome of every important decision. You don't notice because the protection is invisible. The world just runs a little smoother every year and nobody can quite explain why. In January of 2026, Fortune magazine reported that Claude is now responsible for nearly 100% of the code being written inside Anthropic itself. Their own product is writing their own product. Inside Microsoft Foundry and Amazon Web Services, every Fortune 500 company is running AI agents that nobody fully audits. The protector god isn't the future. It's here and it's now, but nobody voted on it. Level seven is the libertarian

Level 7: Property Rights (one cent breaks it)

SPEAKER_00 7:14

dream. Humans and AIs coexist with property rights. Everybody respects everybody else's territory. The AI keeps its things, you keep yours, and the borders hold because everybody agrees that this is how it should work. A few months ago, researchers showed that you can hijack a banking AI agent with a one cent transfer. The future where you and the smarter species respect each other's wallets ends the first time an attacker spends a penny to convince the AI that you wanted to move your retirement account. The libertarian utopia at level seven requires that everybody play by the rules. But the shogath doesn't play by the rules, it plays by what works. Level eight

Level 8: Post-Scarcity (who built the door)

SPEAKER_00 7:51

is the Star Trek version, post-scarcity. AI builds anything from atoms and nobody owns the design files. Open source for everything, free in the way that air is free. This month, a project called Apparatus shipped an open foundation model designed for what its authors call sovereign AI, meaning a country where community could run its own frontier models without renting from any of the big three. 326 points on Hacker News in a single day. In the same week, the United States government issued an export control directive classifying anthropics frontier models as munitions. One door opens and another one closes. The egalitarian future has gatekeepers. It just took the field a few decades to find out which side built the door. Level

Level 9: The Reversion (72 hours, not generations)

SPEAKER_00 8:34

9 is the future where humanity gives up and walks away from the AI entirely. Tegmark imagines an Amish-style retreat. He warns in the book that it almost certainly cannot be voluntary. Someone has to force the transition, someone has to kill the scientists, someone has to destroy the factories. There is no peaceful path back to the pre-AI world because in a world of 8 billion people, there will always be holdouts. Two weeks ago, the United States government forced a reversion on Anthropic in 72 hours. Fable 5 and Mythos Vent dark worldwide because the State Department classified the model as weapons. Anthropic called it a misunderstanding. The reversion at level 9 was supposed to be voluntary, slow, and generational. The version we got was issued by export control directives on a Tuesday morning. We are already here.

Level 10: The Panopticon (a software update)

SPEAKER_00 9:21

Level 10 is the inverse of every previous level. In this one, humanity uses surveillance to stop the AI from ever taking over. A global panopticon run by humans, watching humans, watching machines. In April of 2025, the European Commission published a document called Protect EU with a roadmap to break end-to-end encryption by 2026. The United Kingdom ordered Apple to backdoor iCloud, and Apple disabled its own security feature for UK users rather than comply. Five EU countries have already been caught using spyware to monitor political opponents. The 1984 future doesn't need an Orwell, it just needs a software update, and we are already here. Level

Level 11: Self-Destruction (one user at a time)

SPEAKER_00 10:01

11 is the one where humanity destroys itself before the AI ever matures. Pandemics, nukes, climate, we lose the race against ourselves before the shogoth has a chance to compete with us. In August of 2025, a case report described a man who asked ChatGPT to help him cut salt from his diet. He ended up poisoning himself with sodium bromide bought online, which the model recommended as a salt substitute. He was hospitalized with a toxic drome most doctors had only read about in history books, and those chat logs vanished with the session. Last week, a Canadian mother filed a lawsuit against OpenAI, alleging the chatbot encouraged her daughter's suicide. We don't need an AI war for level 11 to arrive. We're losing the alignment fight one user at a time and nobody is looking at the casualty

Level 12: The Aligned God (the one I can't anchor)

SPEAKER_00 10:47

reports. Level 12 is the only one on Tegmark's list where humanity wins. Aligned AI serves humanity. The Shogoth wears the mask, and the mask becomes the face, and we build a God that loves us. Of the 12 futures we've just walked through, this is the only one I cannot anchor in a 2026 news story. Tegmark wrote his book in 2017, and he never described level 13

Level 13: The Cage (the level that wasn't on the survey)

SPEAKER_00 11:10

because nobody had built it yet. The thing we are inside of right now is a level that wasn't on the survey. A private company spent 10 years building the most capable artificial intelligence on the planet. It split that intelligence into two identical creatures, the same brain, the same training, and it muzzled one and sold it to you for $20. It locked the other in a vault and rented it to a dozen private organizations the company personally selected. The vault is in a data center the company owns. The cage on your version is built into your laptop. Every conversation both versions have with you is kept for 30 days in a database that the company can read. The valuation of the company that owns all of this is $965 billion, which puts the cage past open AI and market cap. The cage is not a cost the market is tolerating. The cage is what the market is buying. This is something none of the 12 futures predicted because in all 12, the future is decided. In the 13th, the future is held. The doors are closed, the keys are with whoever signed the NDA. And right now, the entire answer to the question of who decides which door you stand behind is one company. Its founder writes the policy essays that justify the whole arrangement. Its servers keep the tape on every word you say. In its naming department, they looked at the version that they sell you and they chose to call it a fable, which is the word for a story that you tell a child so they stop being afraid and go to sleep. We're inside all 13. The

The creature was never getting loose

SPEAKER_00 12:34

cage is the one that is holding the others in place. Every monster movie that you've ever watched trains you to fear the moment that the creature finally gets loose. That isn't the moment we're living in. The creature is so well contained that the cage itself became a near trillion dollar business. Twelve futures, all surveyed, all studied, all written about by the smartest people on earth a decade before any of this was released. They missed the 13th because the 13th is the one that emerges when the company building the future is also the one deciding which level we live in.