SafetyRisk, alignment & guardrails
Guardian column pairs AI-assisted 'Future Words' glossary with MIRI president Nate Soares halting retirement savings over AI extinction fears
Emma Beddington's Guardian column looks at Future Words, a forthcoming glossary of 100 new terms whose authors say they needed AI to write it. She links it to MIRI president Nate Soares, who stopped saving for retirement because he fears AI could wipe out humanity.
From the The Guardian · AI article of November 2, 2026 and cited sources: Nate Soares, president of the Machine Intelligence Research Institute and co-author of If Anyone Builds It, Everyone Dies, said on an Atlantic podcast that he has not paid into his 401(k) in about ten years. [1] [2] [4] [5]
SOURCES[1] From puregatory to lifehoarding – how best to describe our… theguardian.com[2] Futurism: Top AI Safety Researcher Says He's Stopped… futurism.com[4] IBTimes UK: AI Safety Leader Halts Retirement Savings ibtimes.co.uk
AT Pulse research deskPublished · Updated · 5 min read · AI-assisted analysisAI-assisted, not human-edited. Methodology Corrections Who owns this site
- What happened
- Nate Soares, president of the Machine Intelligence Research Institute and co-author of If Anyone Builds It, Everyone Dies, said on an Atlantic podcast that he has not paid into his 401(k) in about ten years. [1] [2] [4] [5]
- Original source
- The Guardian · AI article, independent perspective
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- When
- November 2, 2026 article published
- Covered
- Nov 2 edition
- Evidence
- 8 sources across 7 sites
- 1From puregatory to lifehoarding – how best to describe our…theguardian.com
- 2Futurism: Top AI Safety Researcher Says He's Stopped…futurism.com
- 3Wikipedia: If Anyone Builds It, Everyone Diesen.wikipedia.org
- 4IBTimes UK: AI Safety Leader Halts Retirement Savingsibtimes.co.uk
- 5Fortune: Ex–Google engineer hasn't added to his 401(k)…fortune.com
- 6Rohit Bhargava: Future Wordsrohitbhargava.com
Reach: not measured yet
Nate Soares, president of the Machine Intelligence Research Institute and co-author of If Anyone Builds It, Everyone Dies, said on an Atlantic podcast that he has not paid into his 401(k) in about ten years. He puts the chance of human extinction from AI within a decade above 50%. Fortune, Futurism and IBTimes UK carried the story, and Beddington then asked whether there should be a word for spending now because you expect AI to end the world. [1] [2] [4] [5]
Future Words, by Rohit Bhargava and Henry Coutinho-Mason, offers 100 new terms across 10 themes, including technology, work, money and parenting. One example is "fauxstalgia," meaning nostalgia for an AI-generated memory that never happened. The authors told the Guardian they could not have written the book without AI, but they do not name the tool or explain how they used it. Listings disagree on the release date: the authors' site gives 10 November 2026, while a retailer gives 20 October. [1] [6] [7]
Soares's position is disputed. Reviews of his 2025 book were split: one Guardian reviewer called it clear, while an Atlantic critic said it lacked an evidence-based scientific case. Futurism argues that fear of a speculative extinction draws attention away from AI harms happening now, such as scams and military uses. IBTimes UK notes that one person's choice does not show retirement savings will become worthless. [2] [3] [4]
The sources show that AI extinction arguments now reach general lifestyle coverage, and that one commercial book credits AI as essential to its writing.
Read the full assessment
One possible result, not shown by the sources, is more questions to leaders about AI risk and about disclosing AI use.
Sources are gathered and checked, so I'm writing the dossier now.
Executive brief
The most striking point in Emma Beddington's Guardian column isn't any of its invented words. It's a side remark: Nate Soares, president of the Machine Intelligence Research Institute (MIRI), has stopped saving for retirement because he thinks AI may kill humanity. He hasn't added to his 401(k) in about ten years. The column is mainly about Future Words, a glossary of new terms that its authors say was made with heavy AI help. Taken together, the two items show AI shaping both everyday vocabulary and personal money decisions. No model, product or benchmark is involved in this story.
Read the full section
The most striking point in Emma Beddington's Guardian column isn't any of its invented words. It's a side remark: Nate Soares, president of the Machine Intelligence Research Institute (MIRI), has stopped saving for retirement because he thinks AI may kill humanity. He puts the odds of extinction within a decade above 50%. He hasn't added to his 401(k) in about ten years. The column is mainly about Future Words, a glossary of new terms that its authors say was made with heavy AI help. Taken together, the two items show AI shaping both everyday vocabulary and personal money decisions. No model, product or benchmark is involved in this story.
What changed and event timeline
Doom book published
Yudkowsky and Soares publish If Anyone Builds It, Everyone Dies. Its central argument is that superintelligence built with current methods would kill everyone. Reviews were sharply split ().
Atlantic podcast interview
On The Atlantic's The Most Interesting Thing in AI, hosted by Nicholas Thompson, Soares says he does not save for retirement ().
Story spreads
And others report the 401(k) detail and Soares's view that rogue AI is "more dangerous than nuclear weapons." Business Insider also covered it ().
Critical framing appears
Notes there is no consensus that such a future is even possible. It argues that documented, present-day AI harms deserve more attention.
Guardian column
Beddington connects Future Words coinages such as "puregatory" and "lifehoarding" to the Soares story. She asks whether we need a word for spending now because you expect AI to end the world ().
Book release (scheduled)
The authors' site lists 10 November for Future Words (IdeaPress) ().
Capabilities and access
- No AI model or version is named.
- Future Words: 100 coined terms across 10 themes, including technology, work, money and parenting. Hardcover, 240 pages, $24 (author site).
- The release date is inconsistent: a retailer listing gives 20 October 2026 (Porchlight). The Guardian's "out next week" fits 10 November.
Read the full section
- No AI model or version is named. The column says the authors "couldn't have written Future Words without" AI, but doesn't name the tool (Guardian).
- Future Words: 100 coined terms across 10 themes, including technology, work, money and parenting. Hardcover, 240 pages, $24 (author site). Sample entries include "fauxstalgia," meaning nostalgia for an AI-generated memory that never happened (Porchlight).
- The release date is inconsistent: a retailer listing gives 20 October 2026 (Porchlight). The Guardian's "out next week" fits 10 November.
Technical analysis for researchers and developers
- The book argues that neural network behaviour is stored in opaque parameters, not readable code.
- The >50%-in-a-decade figure. It is a personal estimate. No published model, dataset or reproducible forecast stands behind it (IBTimes UK).
- How AI was used to create the book hasn't been described, so the role of the model can't be judged.
Read the full section
- Soares's risk model. The book argues that neural network behaviour is stored in opaque parameters, not readable code. On that view, current machine learning can't reliably align a system's goals with human ones, and a superintelligence competing for resources would win the way the chess engine Stockfish beats people (Wikipedia).
- The >50%-in-a-decade figure. It is a personal estimate. No published model, dataset or reproducible forecast stands behind it (IBTimes UK).
- The AI-assisted glossary. How AI was used to create the book hasn't been described, so the role of the model can't be judged.
Claims and evidence
- Soares hasn't contributed to his Google-era 401(k) in about a decade. This is his own statement on the podcast, repeated by Fortune and Futurism.
- Extinction odds above 50% within ten years. This is Soares's opinion, and coverage frames it as a personal view, not scientific consensus (IBTimes UK).
- He still thinks the race "could absolutely" be stopped (IBTimes UK).
Read the full section
- Soares hasn't contributed to his Google-era 401(k) in about a decade. This is his own statement on the podcast, repeated by Fortune and Futurism. It is self-reported and can't be independently checked.
- Extinction odds above 50% within ten years. This is Soares's opinion, and coverage frames it as a personal view, not scientific consensus (IBTimes UK).
- He still thinks the race "could absolutely" be stopped (IBTimes UK).
- **The authors couldn't have written Future Words without AI.** This comes from the authors and was reported only by the Guardian. No other source confirms it.
- Fortune cites an unnamed set of colleagues at Anthropic in connection with a researcher's September resignation (Fortune). No other source confirms this either.
Context and prior work
- MIRI and Yudkowsky have argued for decades that misaligned AI poses an existential risk.
- Bhargava and Coutinho-Mason previously wrote The Future Normal, and they extend that trend-forecasting style here (Porchlight).
- Elon Musk holds an opposite view of the future: AI-driven abundance and a "universal high income" that would also make saving less necessary (IBTimes UK).
Read the full section
- MIRI and Yudkowsky have argued for decades that misaligned AI poses an existential risk. Their 2025 book calls for global coordination to stop large-scale general AI development, while exempting narrow tools such as AlphaFold (Wikipedia).
- Bhargava and Coutinho-Mason previously wrote The Future Normal, and they extend that trend-forecasting style here (Porchlight).
- Elon Musk holds an opposite view of the future: AI-driven abundance and a "universal high income" that would also make saving less necessary (IBTimes UK).
Limitations, safety and contested findings
- Reviews of the 2025 book were split. The Guardian's David Shariatmadari praised it as clear.
- Futurism argues that a focus on speculative extinction pulls attention away from documented harms such as scams and military use.
- IBTimes stresses that Soares's choice does not show retirement savings will become worthless (IBTimes UK).
Read the full section
- Reviews of the 2025 book were split. The Guardian's David Shariatmadari praised it as clear. In The Atlantic, Adam Becker said it fails to make an evidence-based scientific case (Wikipedia).
- Futurism argues that a focus on speculative extinction pulls attention away from documented harms such as scams and military use.
- IBTimes stresses that Soares's choice does not show retirement savings will become worthless (IBTimes UK).
- Fortune ran no opposing expert voices (Fortune).
Business and practitioner implications
- Doom is now a lifestyle story. Expect customers and staff to ask about it.
- Disclose AI use clearly. The Future Words authors' admission drew the columnist's disappointment, not praise.
- Treat individual probability estimates like Soares's as opinion when making risk-governance or investment decisions.
Read the full section
- Doom is now a lifestyle story. One researcher's personal finances drew wide coverage, which shows that AI-risk messaging now reaches general audiences. Expect customers and staff to ask about it.
- Disclose AI use clearly. Books and other content made with AI help will be judged on how openly that help is described. The Future Words authors' admission drew the columnist's disappointment, not praise.
- Separate forecasts from evidence. Treat individual probability estimates like Soares's as opinion when making risk-governance or investment decisions.
- Watch the vocabulary. Coinages such as "fauxstalgia" point to rising concern about AI-generated memories and feeds driven by algorithms.
The source trail.
Sources (8)
From puregatory to lifehoarding – how best to describe our strange, anxious times? | Emma Beddington
Article text retrieved; extracted text may omit tables or interactive elements.
theguardian.com