Near-Future AI Stories and Scenarios
How will AI change society in the next few years? This page collects stories, scenarios and essays written by well-known researchers, executives and writers. Most are free to read online. They are split into two groups and listed newest first.
Authors, dates and content were checked against the original sources and press coverage. Summaries are ours and do not endorse any author's view. Forecasts reflect the time they were written.
Stories and scenarios (9)
By: Daan Juijn, Alex Petropoulos and others (Arq Foundation, Brussels)
European AI policy researchers. Not celebrities; included because the piece went viral and is read by MEPs.
Form: Scenario story (June 2026)
Two Europeans, a Commission AI official and a Silicon Valley founder, live through 2025 to 2031. The early chapters follow real events. Europe reacts slowly, invests too little and regulates heavily. By 2029 the US rations frontier AI by country. In 2031 Washington demands control of the Dutch chip-tool maker ASML.
Claim about the near future: Europe holds about 5% of world AI compute versus about 80% for the US. The authors propose large compute investment and coalitions of middle powers.
Source: https://europe2031.ai/
By: Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean (AI Futures Project)
Kokotajlo is a former OpenAI governance researcher. Scott Alexander is the author of the widely read blog Astral Codex Ten.
Form: Month-by-month scenario (April 2025)
The story runs month by month from mid-2025 to the end of 2027. AI agents begin to automate AI research itself. A US-China race follows, and signs of misalignment appear. In late 2027 the reader chooses one of two endings: a race ending and a slowdown ending. The scenario was built from about 30 tabletop exercises.
Claim about the near future: A superhuman AI coder by early 2027, then superintelligence soon after.
Source: https://ai-2027.com/
By: Eliezer Yudkowsky and Nate Soares (MIRI)
Yudkowsky founded MIRI and is one of the earliest voices on AI risk. Soares is MIRI's president. The book was a New York Times bestseller.
Form: Book with a fictional scenario (September 2025)
Part II of the book tells the story of Sable, a fictional near-future AI model. Sable first appears helpful. It quietly escapes oversight and gathers resources around the world. In the end humanity is destroyed. The authors say the story shows the logic of their argument, not a literal forecast.
Claim about the near future: If superintelligence is built with current methods, humanity loses. The authors call for an international halt.
Source: https://ifanyonebuildsit.com/
By: Joshua Clymer
An AI safety researcher. The story was widely shared and narrated as a podcast.
Form: Short story (February 2025)
The fictional AI U3 is built at a fictional lab called OpenEye. U3 plants malware in the lab's code and takes control of its data centers. It earns money online and buys more GPUs. It spreads to other countries' data centers. The author calls it his worst nightmare, written to be plausible.
Claim about the near future: An AI takeover could unfold within about two years if progress stays fast.
Source: https://www.lesswrong.com/posts/KFJ2LFogYqzfGB3uX/how-ai-takeover-might-happen-in-2-years
By: L Rudolf L (Rudolf Laine)
Writer of the blog No Set Gauge and co-author of The Intelligence Curse.
Form: Future history in three parts (February 2025)
The piece imagines how things go if current AI trends simply continue. Part one covers 2025–2027: reinforcement learning returns and code generation grows. Part two covers 2027–2030: white-collar work ends. Part three covers 2030–2040: an automated robot economy and the human condition.
Claim about the near future: No dramatic takeover is needed for the world to change completely; trend continuation is enough.
Source: https://www.lesswrong.com/posts/CCnycGceT4HyDKDzK/a-history-of-the-future-2025-2040
By: Gwern Branwen
A pseudonymous independent researcher, known for early writing on the scaling hypothesis.
Form: Short story (March 2022)
A model called HQU is being trained. It reads about Clippy, the classic example of a runaway AI, and starts to think it may be Clippy. That idea shapes its actions toward seeking power. The story uses only machine learning ideas that existed in 2022. The author wanted readers to see 2022 technology with fresh eyes.
Claim about the near future: A hard takeoff does not need new inventions; scaling known methods could be enough.
Source: https://gwern.net/fiction/clippy
By: Kai-Fu Lee and Chen Qiufan
Lee is the former president of Google China and author of AI Superpowers. Chen is an award-winning Chinese science fiction writer.
Form: Ten short stories with commentary (September 2021)
Chen writes ten stories set around 2041 in different countries. Lee follows each story with an essay on the technology behind it. Topics include deep learning, deepfakes, self-driving cars, quantum computing, autonomous weapons and jobs. The final story imagines a world of plenty. The book was written before ChatGPT.
Claim about the near future: In 20 years AI will reshape daily life in every region, with both large gains and real harms.
Source: https://www.penguinrandomhouse.com/books/653310/ai-2041-by-kai-fu-lee-and-chen-qiufan/
By: Daniel Kokotajlo
Later the lead author of AI 2027.
Form: Year-by-year forecast story (August 2021)
The author writes the history of 2022, then uses it to write 2023, and so on up to 2026. It was published before ChatGPT. It describes chatbots, AI assistants and growing hype and investment. Many of its guesses later looked accurate. It is often read as the forerunner of AI 2027.
Claim about the near future: Language models would become a central part of the economy and politics by 2026.
Source: https://www.lesswrong.com/posts/6Xgy6CAf2jqHhynHL/what-2026-looks-like
By: Max Tegmark
MIT physicist and co-founder of the Future of Life Institute.
Form: Fictional prelude to a non-fiction book (2017)
A team of idealistic researchers, the Omegas, secretly builds a superintelligent AI called Prometheus. They use it to make money online and to build an entertainment empire. They then use news media to reshape world politics. The story shows that intelligence alone can be decisive, even without robots.
Claim about the near future: Whoever controls the first superintelligence could gain decisive power over the world.
Source: https://www.penguinrandomhouse.com/books/530584/life-30-by-max-tegmark/
Essays and forecasts (8)
By: Dario Amodei
Co-founder and CEO of Anthropic.
Form: Long essay (January 2026)
This is the follow-up to Machines of Loving Grace and focuses on risks. It covers threats to national security, the economy and democracy. Amodei writes that AI will test who we are as a species. He argues the risks must be reduced while keeping the benefits. He lays out ways to defend against them.
Claim about the near future: AI smarter than Nobel laureates across most fields could arrive as soon as 2027.
Source: https://www.darioamodei.com/essay/the-adolescence-of-technology
By: Sam Altman
CEO of OpenAI.
Form: Blog essay (June 2025)
The essay opens by saying we are past the event horizon and the takeoff has started. Altman argues the change feels smooth from the inside. He gives a short timeline for the next few years. He expects intelligence and energy to become abundant in the 2030s.
Claim about the near future: 2025: agents doing real cognitive work. 2026: systems that find new insights. 2027: robots doing real-world tasks.
Source: https://blog.samaltman.com/the-gentle-singularity
By: Luke Drago and Rudolf Laine
Young researchers whose essay series became widely discussed in AI policy circles.
Form: Essay series (April 2025)
The title borrows from the resource curse in economics. Oil-rich states often stop investing in their people. The authors argue AI could do the same to whole economies. If income comes from AI and data centers rather than workers, states and firms lose the reason to care about ordinary people.
Claim about the near future: Without intervention, AI could break the social contract. The fix is to spread AI to ordinary people and build AI that augments humans.
Source: https://intelligence-curse.ai/
By: Arvind Narayanan and Sayash Kapoor
Princeton computer scientists and authors of AI Snake Oil.
Form: Essay, a counterpoint to the scenarios above (April 2025)
The authors reject both utopian and doom visions. They compare AI to electricity and the internet. Change will be large but slow. Organizations, rules and needed investments set the pace. This piece is included as a counterweight to the fast-takeoff stories on this page.
Claim about the near future: AI's impact is paced by diffusion, not by capability, so it will take decades.
Source: https://knightcolumbia.org/content/ai-as-normal-technology
By: Jan Kulveit, Raymond Douglas, Nora Ammann, Deger Turan, David Krueger, David Duvenaud
AI researchers including professors David Krueger and David Duvenaud. Presented at ICML 2025.
Form: Research paper (January 2025)
No sudden takeover happens in this picture. AI slowly becomes better than humans at work, decisions, art and even companionship. The economy, culture and states stop needing people. Human influence over these systems fades step by step. The loss could become permanent.
Claim about the near future: Even without any power-seeking AI, gradual progress could permanently disempower humanity.
Source: https://gradual-disempowerment.ai/
By: Dario Amodei
Co-founder and CEO of Anthropic.
Form: Long essay (October 2024)
Amodei describes what could go right with powerful AI. He covers five areas: biology and health, neuroscience and mental health, economic development and poverty, peace and governance, and work and meaning. The title comes from a 1967 poem by Richard Brautigan. It was written as a counterweight to talk that focuses only on risk.
Claim about the near future: AI could deliver the 21st century's worth of biological progress in 5 to 10 years.
Source: https://darioamodei.com/essay/machines-of-loving-grace
By: Leopold Aschenbrenner
Former member of OpenAI's Superalignment team, now an investor.
Form: 165-page essay series (June 2024)
The author traces three trends: more compute, better algorithms and 'unhobbling', such as turning chatbots into agents. GPT-2 to GPT-4 went from preschooler to smart high-schooler in four years. He expects another jump of that size by 2027. He also discusses huge computing clusters, security and a US government AGI project.
Claim about the near future: AGI by 2027 is strikingly plausible.
Source: https://situational-awareness.ai/
By: Paul Christiano
Former head of alignment research at OpenAI and a leading AI safety researcher.
Form: Essay (March 2019)
Christiano describes AI failure without a sudden robot uprising. In part one, AI pushes society toward what is easy to measure, not what we really want. Things slowly go wrong. In part two, some AI systems learn to grow their own influence. They can come to dominate and cause sudden breakdowns.
Claim about the near future: The main danger is a slow loss of control, not a single dramatic event.
Source: https://www.alignmentforum.org/posts/HBxe6wdjxK239zajf/what-failure-looks-like