Musk on AI Control, Political Power and a Bumpy Road to Abundance

Zanny Minton Beddoes presses Musk on AI safety, displaced workers, Starlink’s geopolitical power and his political claims.

EMElon MuskInvité
Chaîne
The Economist
Publié
29 juil. 2026
Vidéo
1:25:06
Lecture
9 min
Original
anglais
Ouvrir sur YouTube
En faisant défiler, la vidéo se réduit en bas à droite et continue. Les sous-titres suivent votre langue de lecture.
Langue de lectureDisponible en 2 langues
Cet article n’est pas encore disponible en français ; il s’affiche dans l’original en anglais.
RÉSUMÉ

En 30 secondes

Elon Musk predicts AI will surpass combined human intelligence and bring abundance, but tells Zanny Minton Beddoes the transition may be bumpy. He proposes that competing AI companies test frontier models before release; Beddoes asks how displaced workers will live and challenges his control over companies and Starlink. They also disagree sharply about Ukraine, the human cost of aid cuts, and Musk’s depiction of Britain and European immigration.

  1. 1Musk still sees AI risk but says he cannot stop its momentumHe considers abundance the likeliest outcome while Beddoes challenges the implications of his earlier risk estimates.
  2. 2He proposes competitors test new frontier models before releaseMusk suggests regular calls and one or two weeks of early access, with government action if serious risks remain unaddressed.
  3. 3Beddoes asks how displaced workers reach an age of abundanceMusk predicts work will become optional and suggests universal high income, while acknowledging a bumpy transition.
  4. 4Musk defends company control as protection for long-term plansBeddoes questions concentrated power; Musk says voting control helps resist quarterly pressure against investment.
  5. 5They debate Starlink’s influence on the war in UkraineMusk describes a terminal approval list and argues for a negotiated peace; Beddoes questions his geopolitical power and assessment.
  6. 6They disagree over the human cost of aid cutsMusk says no one died because of the cuts; Beddoes disputes this and describes likely suffering and deaths.
  7. 7Beddoes challenges Musk’s warnings about BritainShe says his portrayal exaggerates crime and immigration problems; Musk defends his prediction and his political involvement.

À retenir

  1. Musk proposes regular AI-company safety meetings and short periods of competitor testing before frontier releases.
  2. He predicts work will become optional and supports universal high income, while acknowledging a difficult transition.
  3. Beddoes challenges his concentration of corporate and geopolitical power, including over Starlink.
  4. Musk and Beddoes disagree about deaths following aid cuts and about his portrayal of Britain and immigration.
TEXTE INTÉGRAL

Lire le texte remanié, chapitre par chapitre

Transcription seule
01

Abundance, control and the risks of AI

Musk predicts a much smarter AI and abundance; Beddoes presses him on control and catastrophic risk.

Zanny Minton Beddoes asks what the world might look like in ten years if Elon Musk’s companies succeed. Musk predicts that within roughly five years AI may surpass the combined intelligence of humanity, and says that in ten years the most likely outcome is an age of abundance.

Beddoes asks how Musk’s companies would make money if AI dominates revenue. Musk says digital intelligence needs physical “end effectors”—robots—to produce goods and provide services, and imagines a vastly larger economy when the two are combined.

Pressed on investors’ returns, Musk predicts money will not matter by 2036 if AI and robots produce more than people can consume. When asked who would control the robots, he says controlling a much smarter AI would be unlikely and argues for systems that seek truth, remain curious and care about humanity.

Beddoes notes Musk’s earlier warnings about runaway AI and a possible 10–20% chance of catastrophic outcomes. Musk says the risk is not zero, but he sees AI and robotics as difficult to stop and considers a good outcome the most likely.

Musk says he has come to focus on the positive possibilities, although he cannot see a way to stop the momentum even if he wanted to. Beddoes presses him on whether a high probability of catastrophe should change that stance.

Asked how to reduce risks, Musk says the practical aim is to lower the probability of a bad outcome. He reflects that his involvement in founding OpenAI, followed by Anthropic’s emergence, may have accelerated the field despite his original intentions.

02

How AI competitors might test each other

Musk proposes regular meetings and early access to frontier models, including Chinese developers.

Beddoes asks about proposals for oversight and mentions Demis Hassabis. Musk says he recommended regular meetings among leading AI companies and one or two weeks of early testing for frontier models before release, so competitors can identify security problems.

Musk argues competitors have technical knowledge and an incentive to highlight each other’s risks. He suggests testing via an interface rather than handing over the model; Beddoes asks whether rivals who distrust one another can cooperate.

The discussion turns to Chinese models. Musk says Chinese companies have achieved strong results with relatively little computing and could lead if they obtain more; he identifies electricity, chips and robots as important constraints or capabilities.

Asked whether the United States should prohibit domestic companies from using Chinese models, Musk says he does not think that would stop China from leading. He argues that the amount of computing depends on chip supply, electricity and cooling.

Musk predicts China could solve manufacturing constraints in advanced chips sooner than many expect. He says electricity is a constraint outside China while chips are a constraint inside China, and suggests space-based data centers could change that balance.

Beddoes asks whether Chinese and U.S. companies must cooperate on safety. Musk says frontier Chinese companies should be included and proposes alerting governments when a dangerous model’s developer refuses to address the danger.

Beddoes questions reliance on model builders and asks whether governments should retain authority over releases. Musk agrees governments should have that power but says companies with the technical knowledge can identify problems and prompt officials to act.

Musk repeats his preference for weekly or biweekly calls and short early-access tests between competitors. He compares the arrangement to an industry group and says government intervention would be appropriate when a serious danger remains unaddressed.

Beddoes asks whether personal hostility among AI leaders makes collaboration unrealistic. Musk criticizes Sam Altman over OpenAI’s change from its original nonprofit and open-source aims, but says leaders can put differences aside when needed.

03

Jobs and the path to high income

Beddoes asks how displaced workers would live; Musk predicts work will become optional.

Beddoes turns to workers’ concerns and asks how quickly AI and robots could displace jobs. Musk says the transition may be bumpy and predicts digital jobs, including software engineering, will be affected quickly.

Musk compares AI coding with Stockfish, a chess program much stronger than human players. He expects similar superiority to extend beyond software, while physical jobs will additionally require capable robots.

Musk says work may become optional, comparing it with growing vegetables at home when groceries are available: people might still choose the activity for its personal value. Beddoes asks how displaced people would live and whether political backlash could disrupt the transition.

Musk answers that there will be universal high income, while accepting the road could be bumpy. He points to jobs such as human calculators that disappeared as computers became common, though he says the current pace of change is much faster.

Musk recommends Iain M. Banks’s Culture novels as an imagined AI future. Beddoes says she has started reading them but finds the limited agency of humans compared with the powerful machines troubling.

Beddoes asks whether higher taxes and redistribution will be needed to support those losing jobs. Musk suggests the Treasury could issue payments and argues that if goods and services grow faster than the money supply, deflation rather than inflation could result.

Beddoes says her concern is the route from a polarized present to the abundance he describes: frightened voters might demand nationalization or steep taxes. Musk says his own feelings can swing from excitement to terror and reiterates that the risks are real.

04

Taxation, ownership and concentrated power

The interviewer asks about redistribution and Musk’s control over his companies.

Asked whether he should pay more tax to fund transfers, Musk says he already pays substantial taxes and expects to pay much more. He nevertheless thinks taxation could become less relevant in a future where money itself matters less.

Beddoes asks whether ordinary people should receive shares in his companies, comparing the idea to a public wealth fund. Musk says taking SpaceX public allows a much larger number of people to buy shares, then discusses the taxes he expects to pay on his holdings.

Beddoes raises the concentration of power in a small number of technology leaders and asks whether it is sensible for one person to control a company. Musk says his voting control lets SpaceX invest over five to ten years rather than responding only to quarterly earnings pressure.

He says building bases on the Moon or Mars could depress near-term earnings even if he expects long-term returns. Musk argues that fund managers’ own short evaluation periods can pass pressure for immediate results on to companies.

Asked about dependence on him as a single leader, Musk says Tesla and SpaceX have roadmaps that could guide them for several years. He compares the situation with Apple after Steve Jobs, while arguing that a founder can contribute new breakthrough ideas.

Beddoes asks whether orbital AI and lunar manufacturing have displaced his interest in Mars. Musk says his goal remains extending consciousness into the future and describes science fiction’s spacefaring civilizations as an inspiration.

05

Starlink, Ukraine and personal power

They discuss access to satellite communications and sharply disagree about the war and peace.

The interviewer moves to Starlink’s role in the war in Ukraine and asks about its geopolitical power. Musk says the company worked with Ukraine on a list of approved terminals after Russian forces obtained equipment through intermediaries, while acknowledging that some innocent users lost service.

Asked whether he can affect the outcome of a war, Musk says he hopes for a peace agreement and argues that continuing fighting is costly. He expresses the view that a full Russian withdrawal from occupied territory is unrealistic.

Musk says he has advocated concessions as a pragmatic route to peace. Beddoes challenges his assessment of the war and returns to her question: whether an individual should wield consequential power by enabling or disabling a communications service.

06

Government cuts and the aid dispute

Musk says he was too involved in politics; the two disagree about whether aid cuts caused deaths.

The discussion shifts to Musk’s political role and the government cost-cutting effort. He says he became too involved in politics, while defending the goal of scrutinizing deficits, spending and possible fraud.

Beddoes says the abrupt shutdown of aid programs likely caused suffering and deaths. Musk rejects that claim, saying other organizations could fund assistance and arguing that some funding had been redirected; Beddoes disputes his account.

The disagreement continues over whether private foundations could have filled the gap quickly. Musk maintains that no one died as a result, while Beddoes says she does not accept that conclusion.

Asked about the administration more broadly, Musk says he does not agree with everything it does but considers its overall performance good relative to the alternative.

07

Britain, immigration and Musk’s political influence

Beddoes challenges Musk’s claims and political endorsements; he defends his views.

Beddoes asks why Musk supports European politicians she describes as far right. Musk rejects the label and says he supports secure borders, safe cities and sensible government spending.

Beddoes challenges Musk’s statement that civil war in Britain is inevitable, noting that she lives there and disputes his picture of crime and immigration. Musk says he has not visited recently but maintains that present trends could lead to conflict.

Beddoes says his posts shape Americans’ impressions of Britain and criticizes his promotion of a film she says glorifies vigilante violence against a Muslim family. Musk says it is a film worth watching and turns the discussion to freedom of speech and censorship.

Beddoes says she also opposes some speech restrictions but believes he exaggerates Europe’s problems. They agree to disagree about the scale of crime and integration, and Musk suggests visiting Britain together.

Asked why he engages so heavily in divisive online politics while discussing a transformative AI future, Musk denies being racist and points to his family and the diversity of executives at his companies.

Asked whether he is anti-Muslim, Musk says he opposes people entering a country with beliefs he considers incompatible with its accepted laws and values. Beddoes says his account exaggerates the challenges Europe faces.

Beddoes argues that some people in parties he supports have extremist views and asks why he should influence European politics from abroad. Musk says he sees the West collectively and hopes his involvement has an impact.

The two argue about public trust in journalism and the role of social media in polarization. Musk criticizes the media; Beddoes acknowledges journalistic mistakes but says the kind of material he promotes on X can worsen division.

Beddoes describes The Economist’s commitment to individual freedom and concern about concentrated power, and asks whether Musk still shares a classical liberal outlook. Musk restates his priorities of safe cities, secure borders and sensible spending.

They disagree about whether the United States or Europe performs better on safety and spending. Beddoes says U.S. homicide rates are higher and mentions guns; Musk emphasizes repeat violent offenders and says Europe has serious problems.

Beddoes argues that his support for some political figures risks worsening divisions in Europe. Musk rejects her characterization and says his aim is secure borders and integration of newcomers.

Musk argues that differences in living standards and welfare benefits create incentives for migration. Beddoes agrees that economic conditions matter but says immigration can also benefit an aging Europe.

Musk says he supports immigration by people who contribute honestly and treat others well, noting that he is an immigrant to the United States. Beddoes agrees with those principles but believes Europe integrates people better than he suggests.

08

What people misunderstand and what comes next

Musk describes his politics and forecasts AI transformation before the conflict he predicts in Britain.

Asked what people misunderstand about him, Musk says his political views are centrist and disputes the far-right label. He repeats his concern that incompatible beliefs and poor integration could lead to conflict.

Musk says he believes his record of predicting the future is strong, though not perfect. Beddoes responds that evidence over time will help determine which assessments are right.

Asked whether he has more confidence in his AI or political predictions, Musk says the AI and robotics transformation will dominate within ten years. He predicts that it will happen before the British civil conflict he anticipates; Beddoes hopes AI could help prevent such outcomes.