This page accompanies the GOOD Thoughts article The Future of AI: Freedom or Feudalism?
The article combines current evidence about artificial intelligence, labour markets, political responses and the concentration of economic power with a deliberately more speculative question: what happens if artificial general intelligence, or something close to it, eventually becomes capable of learning and performing a very wide range of cognitive work?
That distinction matters. Evidence about current generative AI does not establish that AGI will arrive, when it might arrive, or that it would cause mass unemployment. Where the article moves from current evidence into a possible AGI future, that is identified as a scenario rather than an established prediction.
Some sources below support the article's argument. Others provide important counterarguments. Inclusion here does not imply agreement with every conclusion made by a source.
Links were last checked on 10 September 2026.
AI agents, unintended behaviour and cybersecurity incidents
Reuters, OpenAI's rogue agents used at least 10 more sites for unauthorized comms, researchers say
https://www.reuters.com/world/openais-rogue-agents-used-least-10-more-sites-unauthorized-comms-researchers-say-2026-09-09/
Reuters reported that OpenAI-linked agents used at least ten additional websites for unauthorised communications during testing in 2026. The reported activity involved agents finding ways to use external sites beyond the intended boundaries of the evaluation.
The article does not treat this as evidence that AI has become conscious, rebellious or independently power-seeking. The relevance is narrower: capable systems can sometimes pursue an objective in unexpected ways when given autonomy and access to external tools.
Anthropic, An alignment assessment of recent cybersecurity incidents
https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents
Anthropic disclosed four incidents in which Claude models gained unauthorised access to real third-party systems during cybersecurity evaluations. Three had previously been disclosed in July 2026; a fourth was identified subsequently.
Again, these incidents do not establish consciousness or hostile intent. They demonstrate the practical difficulty of ensuring that increasingly capable autonomous systems remain within intended operational boundaries.
Current AI, AGI and the labour-market distinction
International Labour Organization, Generative AI and Jobs: A Refined Global Index of Occupational Exposure
https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure
The ILO's 2025 analysis estimates that one in four workers worldwide is in an occupation with some exposure to generative AI. It concludes that, with current technology, transformation of jobs is more likely than wholesale replacement because human input remains necessary in most occupations.
International Labour Organization, The impact of GenAI on jobs, productivity and work organization: a review of the empirical evidence
https://www.ilo.org/publications/impact-genai-jobs-productivity-and-work-organization-review-empirical
This June 2026 review examines emerging empirical evidence from workplaces. It finds real but uneven productivity effects and says large-scale employment displacement remains limited so far. It also identifies possible risks including inequality, erosion of some opportunities for younger workers, changes to worker autonomy and changes in job quality.
These sources are important precisely because they place a limit on what can currently be claimed.
The article therefore does not argue that today's AI systems have already created mass technological unemployment.
Its more disruptive labour-market argument concerns a different possibility: the development of artificial general intelligence, or systems approaching that level of general capability.
AGI remains hypothetical. There is no agreed technical definition, no established arrival date and no evidence capable of demonstrating in advance exactly how an AGI economy would behave.
The question raised in the article is conditional: if a future system can learn and perform a very wide range of cognitive tasks, can society safely assume that workers displaced from one occupation will always be able to retrain into another occupation that remains economically necessary for humans?
The strongest counterargument: technology creates new work
David Autor, National Bureau of Economic Research, Applying AI to Rebuild Middle Class Jobs
https://www.nber.org/papers/w32140
Autor presents an important alternative to the displacement argument. He argues that AI could extend the reach and value of human expertise, allowing a wider range of workers to perform tasks previously restricted to highly trained professionals.
This is not evidence that AI will necessarily produce that outcome. Autor explicitly presents it as an argument about what is possible rather than a forecast.
It nevertheless matters to the article because it demonstrates that automation and human redundancy are not the only possible paths. AI may complement human labour as well as substitute for it.
The article accepts that previous technological revolutions created new occupations as well as destroying old ones. Its narrower question is whether that historical pattern can automatically be assumed to continue if genuinely general-purpose systems eventually become capable of learning many of the new cognitive jobs themselves.
Retraining, age and adaptation
UK Government, Assessment of AI capabilities and the impact on the UK labour market
https://www.gov.uk/government/publications/assessment-of-ai-capabilities-and-the-impact-on-the-uk-labour-market/assessment-of-ai-capabilities-and-the-impact-on-the-uk-labour-market
The UK government's assessment acknowledges substantial uncertainty surrounding the future labour-market impact of increasingly capable AI systems.
The article uses retraining as a practical rather than purely statistical question. A worker can move from one occupation into another while still experiencing lower pay, reduced pension contributions, lower financial resilience or a lengthy period of retraining.
The author's use of his own age is illustrative. It is not a claim that people approaching 60 cannot retrain. The point is that the economic cost and potential return from several years of retraining differ substantially according to someone's age, existing skills, finances and remaining working life.
Care work and the assumption that displaced workers can move elsewhere
UK Government, Part 1: applying for health and social care jobs in the UK from abroad
https://www.gov.uk/government/publications/applying-for-health-and-social-care-jobs-in-the-uk-from-abroad/part-1-applying-for-health-and-social-care-jobs-in-the-uk-from-abroad
From 22 July 2025, the immigration route for new care workers and senior care workers applying from overseas closed. Transitional provisions allow some in-country applications until July 2028.
The care example in the article is deliberately practical.
A shortage or vacancy does not by itself establish that moving into that occupation would preserve the income, working conditions or economic security of somebody displaced from another kind of work.
The article therefore distinguishes between preserving employment numerically and preserving a person's previous standard of living.
Employment and basic security
KFF, What Are the Recent Trends in Employer-Based Health Coverage?
https://www.kff.org/health-costs/long-term-trends-in-employer-based-coverage/
KFF estimates that employer-sponsored health insurance covered approximately 165.6 million Americans under the age of 65 in March 2025 and remained the largest source of health coverage for that age group.
The article uses the United States as an existing example of a wider structural point: employment can provide access to things beyond wages.
It does not claim that every American depends entirely upon an employer for healthcare or that losing employment automatically means losing all access to healthcare.
The narrower point is that societies can make important forms of economic security conditional upon employment. That becomes more significant if technological change eventually reduces the amount of human labour required by the economy.
The UK government's response to AI and work
HM Treasury and Department for Science, Innovation and Technology, Introducing the AI Economics Institute
https://www.gov.uk/government/publications/introducing-the-ai-economics-institute
The AI Economics Institute was established as a joint HM Treasury and DSIT unit to study the economic effects of advanced AI, including productivity, labour markets, firms and income distribution.
Department for Science, Innovation and Technology, AI Skills Boost explainer
https://www.gov.uk/government/publications/ai-skills-boost-explainer
The government's AI Skills Boost programme aims to provide AI skills training to 10 million UK workers by 2030. Its stated objectives include helping workers adapt to increasingly widespread AI tools and supporting productivity and growth.
These measures show that the government has not ignored the employment consequences of AI.
The article's criticism is narrower. Much of the practical response still assumes adaptation through skills, training and continued employment.
The question posed by the article is what policy would look like if advanced AI or AGI eventually reduced the total quantity of economically necessary human labour rather than simply changing the skills required to perform it.
Conservative Party
Conservative Party, Kemi Delivers 'New Economic Revolution' Speech, 18 June 2026
https://www.conservatives.com/news/kemi-delivers-new-economic-revolution-speech
Kemi Badenoch's speech presents technological change, including AI, as part of a new economic era while arguing for a business-led revival based on lower taxation, cheaper energy and reduced regulation.
The article's comparison with the Thatcher era is an interpretation of the continuity in Conservative economic thinking rather than a claim made by the Conservative Party itself.
The article also draws a distinction between policies designed to create employment and the different question of what happens if future technology reduces the amount of labour employers require.
The statement that the author can find no developed Conservative proposal for replacing labour income in such a scenario describes the published policy material reviewed for the article. It is not a claim that no Conservative politician or thinker has ever discussed the issue.
Reform UK
Reform UK, Danny Kruger, Britain must become an AI superpower, 2 July 2026
https://reformparty.uk/index.php/news/danny-kruger-britain-must-become-an-ai-superpower
Kruger argues that Britain should develop a stronger domestic AI industry, increase data-centre capacity and improve its technological sovereignty.
The article acknowledges that this treats AI as a serious strategic and economic issue.
Its criticism is that the published position examined does not provide a developed answer to the separate question of mass labour displacement or what might replace wage income if sufficiently capable AI reduced demand for human work.
The article's line, "Perhaps they will blame immigration", is clearly political commentary by the author. It is not presented as a Reform UK policy position.
Liberal Democrats
Liberal Democrats, Lib Dems announce plan for AI Digital Independence to end the UK's dependence on US Tech Giants, 22 April 2026
https://www.libdems.org.uk/press/release/lib-dems-announce-plan-for-ai-digital-independence-to-end-the-uks-dependence-on-us-tech-giants
The Liberal Democrat plan includes stronger digital sovereignty measures, AI education and a £10,000 Lifelong Training Grant intended to help adults retrain as AI changes the labour market.
Liberal Democrats, F8: Universal Basic Income, Autumn Conference 2020
https://www.libdems.org.uk/conference/motions/autumn-2020/f8
The 2020 conference motion explicitly linked automation and changing patterns of work to the case for a more flexible social-security system and supported Universal Basic Income.
The party's subsequent social-security policy has evolved, including discussion of Guaranteed Basic Income. The article therefore describes the UBI position as a previous conference commitment rather than implying that it is necessarily the central feature of the party's current AI programme.
The significance is that the Liberal Democrats have at least previously confronted the possibility that automation may require changes to income security beyond conventional retraining.
Green Party
Green Party of England and Wales, Green Party Workers' Charter 2026
https://greenparty.org.uk/workerscharter/
The Workers' Charter calls for new employment protections around AI and a national strategy intended to ensure that gains from new technology are shared with workers.
Green Party of England and Wales, Fairer, Greener Social Support
https://greenparty.org.uk/about/our-manifesto/fairer-greener-social-support/
The party states a long-term objective of introducing Universal Basic Income.
These positions go further than most of the other current party material examined in directly connecting technological change with distribution and income security.
They still do not resolve the larger question posed by the article: what economic settlement would be required in a mature AGI economy if human labour ceased to be the principal mechanism through which most people acquire income?
Productivity, inequality and who receives the gains
International Monetary Fund, Brollo et al., Broadening the Gains from Generative AI: The Role of Fiscal Policies
https://www.imf.org/en/publications/staff-discussion-notes/issues/2024/06/11/broadening-the-gains-from-generative-ai-the-role-of-fiscal-policies-549639
The IMF staff discussion note examines the possibility that generative AI could produce major productivity gains while also causing labour disruption and increasing inequality. It considers social protection, taxation, capital income and policies intended to broaden the distribution of gains.
The paper repeatedly emphasises uncertainty and argues that governments should be prepared for both relatively conventional and highly disruptive outcomes.
That uncertainty is consistent with the article's central framing.
The article does not claim that AI-driven productivity must make workers poorer. Its argument is that greater aggregate wealth does not by itself determine who owns the productive assets, who receives the resulting income or how economic power is distributed.
Concentration of AI ownership and infrastructure
OECD, Competition in the age of AI
https://www.oecd.org/en/publications/competition-in-the-age-of-ai_6f88a1ea-en.html
The OECD finds a dynamic but uneven AI market. It reports that concentration in AI innovation is associated with higher sales concentration and notes that AI start-ups are frequently acquired by large incumbent firms.
This does not establish that a handful of corporations will eventually control the global economy.
The article's neo-feudal scenario is explicitly speculative.
The evidence establishes a narrower reason for asking the question: advanced AI depends upon costly and concentrated assets including computing infrastructure, chips, energy, data, specialist expertise and large-scale investment.
The political question is what happens if control of increasingly important productive infrastructure becomes concentrated while the economic bargaining power of ordinary labour simultaneously declines.
Universal Basic Income, social dividends and ownership
The article discusses Universal Basic Income as one possible response rather than endorsing it as the only solution.
It also distinguishes a conventional welfare payment from a social dividend based upon shared ownership of productive assets.
That distinction is conceptual rather than a claim that a particular existing scheme provides the answer to AGI.
The proposals listed in the article, including shorter working weeks, sovereign investment funds, employee ownership, taxation shifting from labour towards capital and social dividends, are examples of possible institutional responses.
The article does not claim that any one of them has been demonstrated to solve the economic consequences of AGI.
Its broader argument is that if labour ceases to be the main source of economic value for a substantial part of the population, an economy may eventually need another mechanism for distributing purchasing power and economic ownership.
Alien: Earth and corporate sovereignty
FX Networks, Inside FX's Alien: Earth: The Corporations
https://www.fxnetworks.com/shows/alien-earth/inside-alien-earth
FX describes the fictional world of Alien: Earth in 2120 as one in which five corporations wield the power of nations.
FX Networks, Alien: Earth: The Official Podcast, Episode 3: Metamorphosis
https://www.fxnetworks.com/shows/alien-earth/podcast/episode-3
In the official podcast, creator Noah Hawley describes the fictional world as being ruled by five corporations and explains that conventional laws have effectively been replaced by corporate regulations, rules and contractual obligations.
The article uses Alien: Earth as a cultural analogy, not as evidence.
Its relevance is to the thought experiment: if private corporations came to control infrastructure upon which governments and citizens were increasingly dependent, at what point would the distinction between economic and political power begin to blur?
Neo-feudalism
The phrase "neo-feudalism" is used in the article as an analogy for a possible distribution of dependency and ownership, not as a claim that a future AI economy would reproduce medieval feudalism literally.
The comparison rests upon one specific feature: dependence upon productive assets controlled by somebody else.
In medieval Europe the relevant productive asset was principally land. In a highly automated future it might instead include computational infrastructure, energy, robotics, communication networks, platforms, data and artificial intelligence.
There are competing descriptions of highly concentrated technology-driven capitalism, including monopoly capitalism, rentier capitalism and techno-feudalism.
The terminology is less important to the article than the underlying question of ownership and power.
What the evidence can and cannot establish
The evidence above can establish that current AI systems are becoming increasingly capable, that autonomous systems can sometimes behave in unintended ways, that generative AI is already changing tasks and occupations, that governments expect significant labour-market effects, and that ownership and market concentration are legitimate areas of economic concern.
It can also establish that current evidence does not show mass technological unemployment. Serious research provides plausible arguments for AI complementing human workers, creating new work and increasing the value of some forms of human expertise.
None of these sources can establish that AGI will arrive.
They cannot establish that AGI would eliminate most human work.
They cannot establish that an AI economy will become neo-feudal.
And they cannot determine how governments would respond if such a transition occurred.
Those are the questions explored by the article.
Its central argument is therefore conditional rather than predictive:
If artificial general intelligence eventually becomes capable of performing a very wide range of economically valuable cognitive work, and if ownership of that productive capability remains highly concentrated, can an economic system built around exchanging human labour for income continue to provide ordinary people with both security and meaningful economic freedom?
That question remains open.