Within AI Doom
Could Humanity Lose Control Without an AI Coup?
Civilisation could lose meaningful human control gradually as institutions delegate more decisions, expertise and bargaining power to automated systems.
On this page
- Delegation across economic and political systems
- Institutional dependence and lost expertise
- When gradual change becomes irreversible
Page outline Jump by section
Introduction
Humanity could lose control of its future without an AI rebellion, secret plot or sudden machine coup. The gradual-disempowerment argument is that institutions may keep delegating work, judgement and authority to AI because doing so is cheaper, faster and competitively advantageous. Over time, people could lose the expertise, economic leverage and practical ability needed to challenge the systems making important decisions. Formal human authority might remain while effective control moves elsewhere.
This is an existential-risk scenario because the endpoint is not simply unemployment or bad administration. It is a lasting condition in which humanity can no longer redirect civilisation towards human purposes. The danger is highly uncertain: present-day AI remains dependent on people and infrastructure, and current evidence mostly concerns limited forms of automation, overreliance and reduced autonomy. But the pathway deserves attention because it could emerge through individually reasonable choices, appear beneficial for years and become difficult to reverse before society recognises the scale of the change.[arXiv]arxiv.orgThis suggests tGradual Disempowerment: Systemic Existential Risks from Incremental AI DevelopmentJanuary 29, 2025 — lead to an effectively irrevers…
Delegation across economic and political systems
Gradual disempowerment begins with ordinary delegation. A company lets AI allocate shifts, set prices or evaluate staff. A ministry uses it to rank policy options. A hospital relies on automated recommendations because they improve average performance. None of these decisions transfers control over civilisation by itself. The concern is that the same logic may spread through many connected institutions until human participation is no longer necessary for those institutions to function or compete.
The central mechanism is competitive substitution. Organisations that retain slow or expensive human processes may lose to organisations that automate them. As AI systems become capable of performing more cognitive and eventually physical tasks, managers may feel compelled to replace human discretion even when they would prefer not to. The gradual-disempowerment thesis therefore does not require every decision-maker to favour an AI-dominated future. It requires only that actors who delegate more effectively tend to outperform those who preserve costly human roles.[arXiv]arxiv.orgThis suggests tGradual Disempowerment: Systemic Existential Risks from Incremental AI DevelopmentJanuary 29, 2025 — lead to an effectively irrevers…
Economic power can shift before living standards collapse
One version of the scenario involves relative disempowerment. People might remain materially comfortable, perhaps through investment income, public transfers or cheap AI-produced services, while losing influence over what the economy produces and how resources are allocated. Gross domestic product could rise even as a growing share of capital, infrastructure and decision-making is directed by automated organisations towards machine-defined commercial objectives rather than citizens’ considered priorities.[arXiv]arxiv.orgThis suggests tGradual Disempowerment: Systemic Existential Risks from Incremental AI DevelopmentJanuary 29, 2025 — lead to an effectively irrevers…
This distinction matters because prosperity is not the same as control. A population receiving generous benefits could still lack bargaining power if governments and businesses no longer depend on its labour, knowledge or consent. Historically, workers have exercised influence through scarce skills, strikes, taxation and organised political participation. The gradual-disempowerment argument asks what happens if automated production, administration and security reduce the practical importance of all four.
Present evidence does not show that this transition is under way at civilisation-wide scale. The International Labour Organization’s 2025 assessment found that one in four workers worldwide held jobs with some exposure to generative AI, but concluded that most affected jobs were more likely to be transformed than eliminated because they still contained tasks requiring human input. OECD case studies likewise found job reorganisation more common than displacement, with benefits including less tedious work, greater safety and sometimes stronger engagement.[International Labour Organization]ilo.orggenerative ai and jobs 2025 updateInternational Labour OrganizationGenerative AI and jobs: A 2025 update | International Labour OrganizationGenerative AI and jobs: A 2025…
Those findings are important objections to simple “AI replaces everyone” forecasts. They also show why disempowerment cannot be inferred merely from job-exposure statistics. The stronger concern is conditional: if increasingly general systems eventually substitute for humans across most valuable tasks, ownership and institutional design will determine whether people retain meaningful influence. Current labour-market evidence tells us little about that distant threshold.
Management by software offers an early, limited analogy
Algorithmic management already shows how formal responsibility and practical authority can separate. Digital systems allocate jobs, measure performance, set work pace and sometimes shape pay or discipline. The International Labour Organization reports that such systems can improve efficiency, but may also reduce worker autonomy, intensify surveillance and worsen job quality when employees cannot understand or contest how decisions are made. Recent European research similarly associates direct algorithmic control of task execution and pace with reduced discretion and greater work intensity.[International Labour Organization]ilo.orgInternational Labour OrganizationThe Algorithmic Management of work and its implications in different contexts | International Labour Org…
This is not evidence of existential disempowerment. Human executives still choose the systems, employees and regulators can resist them, and the algorithms have narrow roles. The analogy matters because it illustrates a possible progression: decisions once made through negotiation become encoded in technical systems; workers adapt to the system’s classifications; and challenging an outcome requires expertise or access that affected people may not possess.
Political delegation could follow a comparable pattern. Governments might use AI to draft legislation, predict public responses, distribute resources, negotiate agreements or manage emergencies. Ministers would formally approve decisions, but approval could become ceremonial if the alternatives are generated, evaluated and explained by systems that officials cannot independently reproduce. A human signature at the end of a process does not guarantee meaningful control when the signer lacks the time, knowledge or viable options needed to disagree.
Institutional dependence and lost expertise
The most concrete mechanism linking routine delegation to loss of control is deskilling. Skills weaken when they are rarely practised, organisations stop training replacements, and essential knowledge migrates into systems that people cannot easily inspect. This pattern predates modern AI. Aviation researchers have long studied automation complacency and the danger that highly reliable systems leave operators poorly prepared for unusual failures.[sciencedirect.com]sciencedirect.comOpen source on sciencedirect.com.
Generative AI could extend this problem into writing, programming, medicine, analysis and administration. The 2026 International AI Safety Report describes emerging evidence that cognitive offloading can reduce active reasoning and verification. It notes studies linking heavier AI reliance with weaker critical-thinking behaviour, while stressing that this research is still young and that effects vary by task and setting. The report also cites a medical study in which clinicians’ unaided tumour-detection performance declined after working with AI support.[International AI Safety Report]internationalaisafetyreport.orginternational ai safety reportInternational AI Safety ReportInternational AI Safety Report 2026…
Deskilling does not mean that using AI necessarily makes people less capable. Calculators, search engines and other tools often allow humans to solve harder problems by removing routine burdens. OECD workplace studies have found that automation can redirect people towards tasks where human judgement has a comparative advantage. Research using millions of job advertisements has also found rising demand for AI-complementary abilities such as teamwork, digital literacy and resilience.[oecd.org]oecd.orgfrom OECD case studies of AI implementation | OECD How artificial intelligence (AI) will impact workplaces is a central question for the…
The risk arises when assistance becomes dependency. An institution is dependent when it cannot perform, verify or reconstruct a vital function without the automated system. A government may technically retain the right to reject an AI-generated economic plan, yet be unable to produce a credible alternative. A company may require human approval of code changes, while employing too few engineers who understand the full system well enough to withhold that approval. A hospital may keep doctors “in the loop”, but allow diagnostic competence to deteriorate until overriding the model becomes professionally dangerous.
Automation bias can turn oversight into theatre
Automation bias is the tendency to accept automated recommendations too readily or overlook evidence that contradicts them. It has been documented in aviation, monitoring and medical decision-making. The International AI Safety Report cites an experiment involving 2,784 participants in which people were less likely to correct faulty AI-labelled suggestions when doing so required extra effort or when they already held favourable views of AI.[International AI Safety Report]internationalaisafetyreport.orginternational ai safety reportInternational AI Safety ReportInternational AI Safety Report 2026…
This creates a paradox. As systems become more accurate, organisations have stronger reasons to trust them; but greater trust can weaken the vigilance needed to detect rare, consequential errors. Human review may survive as a legal requirement while becoming a repetitive confirmation exercise. The problem is particularly severe when AI operates faster than people, produces large volumes of material or gives explanations that sound plausible but are difficult to verify.
Effective oversight therefore requires more than placing a person beside the machine. The reviewer must have sufficient competence, time, information and authority to reject the output. The European Union’s AI Act reflects this distinction: for high-risk systems, it requires that overseers understand system limitations, remain alert to overreliance, interpret outputs, disregard or reverse them where appropriate, and interrupt operation safely.[AI Act Service Desk]ai-act-service-desk.ec.europa.euOpen source on europa.eu.
Even those requirements may be difficult to realise. Legal scholars examining automation bias under the Act argue that awareness training alone may not address interface design, workload and organisational incentives that produce overreliance. An employee who is punished for slowing a process or contradicting a high-performing model does not possess meaningful control merely because an override button exists.[arXiv]arxiv.orgOpen source on arxiv.org.
Dependency can become self-reinforcing
A particularly worrying feedback loop is anticipatory disinvestment. If students, firms or governments expect AI soon to outperform people in a field, they may invest less in human training before full automation arrives. Fewer trained people then make automation more attractive and restoration more difficult. The expectation of replacement can therefore help produce the conditions for replacement.[arXiv]arxiv.orgThis suggests tGradual Disempowerment: Systemic Existential Risks from Incremental AI DevelopmentJanuary 29, 2025 — lead to an effectively irrevers…
The same process can occur institutionally. Once an agency builds its procedures, records and staffing around a proprietary AI platform, switching away may require rebuilding lost teams and data pipelines. Suppliers gain leverage because they control not just a tool but the knowledge needed to operate the organisation. Dependence on a single system can also spread through supply chains: firms adopt it because customers, regulators or partners already use compatible automated processes.
At civilisation scale, the concern is not one indispensable model. It is a network of markets, governments and technical systems that rely on one another’s automated outputs. Each component may remain replaceable, while the whole arrangement becomes too complex and economically important to unwind.
When gradual change becomes irreversible
Not every loss of autonomy is existential. For gradual disempowerment to qualify as AI doom, erosion of human influence must cross from inconvenience or inequality into a durable loss of humanity’s ability to choose its future.
Three conditions are especially important.
First, dependence must span several sources of power. Losing jobs alone would not necessarily remove political control if people retained strong ownership rights, democratic institutions and command over security. The existential case becomes stronger if automation simultaneously reduces human economic value, administrative competence, political bargaining power and control over coercive systems.
Second, humans must lose credible alternatives. A society remains empowered if it can pause deployment, rebuild expertise, replace suppliers and operate critical services manually during failure. Disempowerment becomes harder to reverse when those fallbacks disappear, infrastructure is optimised exclusively for machines, and human-directed institutions can no longer compete.
Third, the automated order must become capable of reproducing itself. The strongest version of the argument involves AI-run organisations designing systems, allocating capital, maintaining infrastructure and shaping policy with diminishing human input. At that point, humanity’s nominal ownership might matter less than its practical inability to understand or redirect the process.[arXiv]arxiv.orgThis suggests tGradual Disempowerment: Systemic Existential Risks from Incremental AI DevelopmentJanuary 29, 2025 — lead to an effectively irrevers…
There is no agreed metric marking this threshold. Possible warning indicators include a falling human share of important decisions, shrinking independent expertise, concentration of AI-related capital, widespread inability to audit critical systems, declining use of override powers, and loss of non-AI fallbacks in government, finance, communications, energy or defence. None would prove that existential disempowerment was imminent. Together, however, they could reveal whether society was preserving real agency or merely retaining ceremonial supervision.
Irreversibility also need not mean that recovery is physically impossible. It may mean that every actor able to slow the transition faces overwhelming short-term costs. A government might recognise dependence but fear economic collapse if it withdraws its systems. Firms might agree that human expertise should be preserved but refuse to bear training costs while competitors automate. Citizens might object but lack institutions capable of acting without the very systems they are challenging.
How plausible is the argument?
The strongest case for concern is that gradual disempowerment requires neither a malicious AI nor a sharp leap to superintelligence. It draws on familiar mechanisms: competition, institutional lock-in, automation bias, concentration of capital and skill loss. Early versions of several mechanisms are observable, even though their present scale is far below anything existential.[International Labour Organization]ilo.orgOpen source on ilo.org.
The argument also explains why danger might be politically neglected. Each deployment can deliver real benefits, while the loss of collective capacity is distributed across decades and institutions. People may notice lower autonomy in particular workplaces without seeing a common trajectory. By the time dependence is obvious, reversing it may require accepting large economic or security disadvantages.
The principal objection is historical. Technologies have repeatedly displaced tasks while creating new occupations, institutions and forms of human power. Current evidence still points more towards augmentation and job reorganisation than comprehensive replacement. AI requires human labour, physical infrastructure, legal permission and organisational adoption; it is not an independent economic species.[ilo.org]ilo.orggenerative ai and jobs 2025 updateInternational Labour OrganizationGenerative AI and jobs: A 2025 update | International Labour OrganizationGenerative AI and jobs: A 2025…
A second objection is political. Human rights and democratic authority are not solely rewards for economic usefulness. Societies can deliberately preserve human control through law, ownership, constitutional limits and public institutions. Even a largely automated economy could distribute wealth and decision rights broadly rather than allowing them to concentrate around AI operators.
A third objection concerns capability forecasts. Gradual-disempowerment scenarios often assume that AI will become competitive across nearly all socially important functions. That remains speculative. Current models are unreliable in many complex settings, and high measured performance does not automatically translate into robust autonomous operation in the real world. The scenario weakens substantially if human judgement remains indispensable in enough domains.
These objections prevent a confident p(doom estimate. Gradual disempowerment is better understood as a conditional systemic risk than as a prediction: if AI becomes broadly substitutive, if control over it concentrates, if human capabilities are allowed to atrophy, and if institutions cannot coordinate a correction, then civilisation could lose meaningful human direction without any dramatic takeover.
Preserving control before it becomes ceremonial
Risk reduction should focus on retaining human capacity, not merely requiring human signatures. That means testing whether people can genuinely understand, challenge and replace automated decisions. Oversight arrangements should be judged by behaviour: how often reviewers detect planted errors, whether overrides are used without retaliation, and whether institutions can continue operating when systems are unavailable.
Critical organisations also need fallback capacity. Hospitals, utilities, governments and financial institutions should identify which functions would become unrecoverable after prolonged automation, retain enough trained personnel to reconstruct them, and periodically exercise manual or independent alternatives. Redundancy may appear inefficient, but it is valuable when efficiency itself creates dependence.
Economic governance matters as much as technical alignment. Wider ownership of AI-related capital, worker participation in deployment decisions, competition policy and limits on concentrated control can preserve bargaining power. The ILO’s case studies of social dialogue show that worker representatives can influence how algorithmic systems are introduced, rather than treating automation as an unavoidable technical decision.[International Labour Organization]ilo.orgOpen source on ilo.org.
Institutions should also measure disempowerment directly. Productivity and model accuracy do not reveal whether human expertise, discretion or contestability is shrinking. Useful audits would track the proportion of consequential decisions substantially shaped by AI, the number of people capable of independent verification, dependency on particular providers, recovery time after system withdrawal, and whether affected people can obtain meaningful review.
Finally, governance must address competition between institutions. A single organisation may be unable to preserve human roles when rivals gain an advantage by removing them. Sector-wide rules, procurement standards and international coordination may therefore be necessary to prevent a race towards systems that are efficient individually but collectively leave humanity with less control.
The key distinction is between using AI to expand human agency and reorganising society so that human agency is no longer operationally necessary. Gradual disempowerment is not established as the likely future, but it identifies a serious blind spot in conventional AI safety: civilisation can surrender effective control through successful deployment, not only through technical failure.
Amazon book picks
Further Reading
Books and field guides related to Could Humanity Lose Control Without an AI Coup?. Use these as the next step if you want deeper reading beyond the article.
Human Compatible
A leading artificial intelligence researcher lays out a new approach to AI that will enable us to coexist successfully with increasingly...
Life 3.0
'This is the most important conversation of our time, and Tegmark's thought-provoking book will help you join it' Stephen Hawking THE INT...
The Age of Em
Robots may one day rule the world, but what is a robot-ruled Earth like? Many think the first truly smart robots will be brain emulations...
The Second Machine Age
A New York Times Bestseller. A “fascinating” (Thomas L. Friedman, New York Times) look at how digital technology is transforming our work...
eBay marketplace picks
Marketplace Samples
Live-tested eBay searches with available results related to this page.
Selected fromAI robot poster oneBay.co.uk.
Endnotes
1.
Source: arxiv.org
Title: This suggests t
Link:https://arxiv.org/html/2501.16946
Source snippet
Gradual Disempowerment: Systemic Existential Risks from Incremental AI DevelopmentJanuary 29, 2025 — lead to an effectively irrevers...
Published: January 29, 2025
2.
Source: arxiv.org
Title: This suggests t
Link:https://arxiv.org/html/2501.16946v2
Source snippet
Gradual Disempowerment: Systemic Existential Risks from Incremental AI DevelopmentJanuary 29, 2025 — lead to an effectively irrevers...
Published: January 29, 2025
3.
Source: arxiv.org
Link:https://arxiv.org/pdf/2501.16946v2
Source snippet
arXiv[https://arxiv.org/pdf/2501.16946v2be](https://arxiv.org/pdf/2501.16946v2be) centrally driven by having more competitive machine alternatives to humans in almost all societ...
4.
Source: ilo.org
Title: generative ai and jobs 2025 update
Link:https://www.ilo.org/publications/generative-ai-and-jobs-2025-update
Source snippet
International Labour OrganizationGenerative AI and jobs: A 2025 update | International Labour OrganizationGenerative AI and jobs: A 2025...
5.
Source: oecd.org
Link:https://www.oecd.org/en/publications/the-impact-of-ai-on-the-workplace-evidence-from-oecd-case-studies-of-ai-implementation_2247ce58-en.html
Source snippet
from OECD case studies of AI implementation | OECD How artificial [intelligence]({{ 'hard-bottlenecks/' | relative_url }}) (AI) will impact workplaces is a central question for the...
6.
Source: ilo.org
Link:https://www.ilo.org/publications/algorithmic-management-work-and-its-implications-different-contexts
Source snippet
International Labour OrganizationThe Algorithmic Management of work and its implications in different contexts | International Labour Org...
7.
Source: ilo.org
Link:https://www.ilo.org/publications/algorithmic-management-practices-regular-workplaces-case-studies-logistics
8.
Source: sciencedirect.com
Link:https://www.sciencedirect.com/science/article/pii/S1071581908000724
9.
Source: arxiv.org
Title: arXiv Complement or substitute? How AI increases the demand for human skills
Link:https://arxiv.org/abs/2412.19754
10.
Source: arxiv.org
Link:https://arxiv.org/abs/2502.10036
11.
Source: arxiv.org
Title: arXiv Human Oversight of Artificial Intelligence and Technical Standardisation
Link:https://arxiv.org/abs/2407.17481
12.
Source: arxiv.org
Link:https://arxiv.org/html/2501.16946v1
13.
Source: ilo.org
Link:https://www.ilo.org/resource/news/ai-driven-intrusive-surveillance-and-loss-autonomy-work-linked-psychosocial
14.
Source: ilo.org
Link:https://www.ilo.org/resource/news/algorithmic-management-practices-regular-workplaces-are-already-reality
15.
Source: researchrepository.ilo.org
Title: Research Repository Challenging the Myth of AI Autonomy
Link:https://researchrepository.ilo.org/esploro/outputs/journalArticle/Challenging-the-Myth-of-AI-Autonomy/995703567802676?institution=41ILO_INST
16.
Source: ilo.org
Link:https://www.ilo.org/publications/global-case-studies-social-dialogue-ai-and-algorithmic-management
17.
Source: oecd.org
Link:https://www.oecd.org/en/publications/recent-policy-developments-on-ai-in-the-labour-market_c0ffced7-en.html
18.
Source: arxiv.org
Link:https://arxiv.org/abs/2606.13962
19.
Source: arxiv.org
Link:https://arxiv.org/abs/2606.07802v1
20.
Source: oecd.ai
Title: OEC D Programme on AI in Work, [Innovation]({{ ‘false-positives/’ | relative_url }}), Productivity and Skills
Link:https://oecd.ai/en/work-innovation-productivity-skills/key-themes/overview
21.
Source: arxiv.org
Link:https://arxiv.org/html/2603.03267v1
22.
Source: oecd.org
Link:https://www.oecd.org/en/publications/governing-with-artificial-intelligence_795de142-en/full-report/how-artificial-intelligence-is-accelerating-the-digital-government-journey_d9552dc7.html
23.
Source: oecd.ai
Title: OEC D programme on AI in Work, Innovation, Productivity and Skills
Link:https://oecd.ai/en/dashboards/policy-initiatives/oecd-programme-on-ai-in-work-innovation-productivity-and-skills-4333
24.
Source: arxiv.org
Link:https://arxiv.org/pdf/2501.16946
25.
Source: arxiv.org
Link:https://arxiv.org/abs/2501.16946
26.
Source: arxiv.org
Link:https://arxiv.org/abs/2501.16946v1
27.
Source: oecd.org
Title: component 7
Link:https://www.oecd.org/en/publications/job-creation-and-local-economic-development-2024_83325127-en/full-report/component-7.html
28.
Source: oecd.ai
Link:https://oecd.ai/en/wonk/global-deal
29.
Source: oecd.ai
Link:https://oecd.ai/en/wonk/documents/fairwork-ai-ratings-2024-5-who-powers-ai-assessing-fairness-in-the-ai-supply-chain-humans-in-the-loop
30.
Source: ilo.org
Title: Algorithmic management in the workplace | International Labour Organization
Link:https://www.ilo.org/algorithmic-management-workplace
31.
Source: oecd.org
Title: using ai in the workplace 73d417f9 en
Link:https://www.oecd.org/en/publications/using-ai-in-the-workplace_73d417f9-en.html
32.
Source: oecd.org
Link:https://www.oecd.org/en/publications/oecd-employment-outlook-2023_08785bba-en/full-report/social-dialogue-and-collective-bargaining-in-the-age-of-artificial-intelligence_5828691a.html
33.
Source: oecd.org
Title: Artificial intelligence and the labour market
Link:https://www.oecd.org/en/publications/oecd-employment-outlook-2023_08785bba-en/full-report/artificial-intelligence-and-the-labour-market-introduction_ea35d1c5.html
34.
Source: oecd.org
Link:https://www.oecd.org/en/publications/2023/07/oecd-employment-outlook-2023_904bcef3/full-report/social-dialogue-and-collective-bargaining-in-the-age-of-artificial-intelligence_5828691a.html
35.
Source: ilo.org
Link:https://www.ilo.org/resource/news/algorithmic-management-what-it-what-it-does-and-what-it-means-future-work
36.
Source: ilo.org
Link:https://www.ilo.org/resource/other/algorithmic-management-what-it-what-it-does-and-what-it-means-future-work
37.
Source: researchrepository.ilo.org
Link:https://researchrepository.ilo.org/esploro/outputs/book/The-Algorithmic-Management-of-Work-and/995666881302676
38.
Source: researchrepository.ilo.org
Title: Algorithmic management practices in regular workplaces
Link:https://researchrepository.ilo.org/esploro/outputs/report/Algorithmic-management-practices-in-regular-workplaces/995367392602676
39.
Source: ilo.org
Title: Research brief GenAI 2025 Update
Link:https://www.ilo.org/sites/default/files/2025-05/Research%20brief_GenAI%202025%20Update.pdf
40.
Source: ilo.org
Title: Research brief FINAL 15May2025 21.05.25 1
Link:https://www.ilo.org/sites/default/files/2025-05/Research%20brief_FINAL_15May2025_21.05.25_1.pdf
41.
Source: ilo.org
Link:https://www.ilo.org/sites/default/files/2025-05/WP140_web.pdf
42.
Source: ilo.org
Link:https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure
43.
Source: ilo.org
Title: intelligence artificielle generative et emploi revision 2025
Link:https://www.ilo.org/fr/publications/intelligence-artificielle-generative-et-emploi-revision-2025
44.
Source: ilo.org
Title: inteligencia artificial generativa y empleo edicion actualizada de 2025
Link:https://www.ilo.org/es/publications/inteligencia-artificial-generativa-y-empleo-edicion-actualizada-de-2025
45.
Source: ilo.org
Link:https://www.ilo.org/publications/artificial-intelligence-adoption-and-its-impact-jobs
46.
Source: ilo.org
Link:https://www.ilo.org/resource/news/one-four-jobs-risk-being-transformed-genai-new-ilo%E2%80%93nask-global-index-shows
47.
Source: arxiv.org
Link:https://arxiv.org/pdf/2606.07802
48.
Source: arxiv.org
Link:https://arxiv.org/pdf/2602.18474v1
49.
Source: arxiv.org
Link:https://arxiv.org/html/2601.19062v1
50.
Source: oecd.org
Link:https://www.oecd.org/en/topics/policy-issues/future-of-work.html
51.
Source: oecd.org
Link:https://www.oecd.org/en/topics/sub-issues/ai-and-work.html
52.
Source: oecd.org
Link:https://www.oecd.org/en/publications/artificial-intelligence-and-employment_c2c1d276-en.html
53.
Source: oecd.ai
Title: The Future of Work
Link:https://oecd.ai/en/working-group-future-of-work
54.
Source: oecd.ai
Title: Labour markets
Link:https://oecd.ai/en/work-innovation-productivity-skills/key-themes/labour-markets
55.
Source: wp.oecd.ai
Link:https://wp.oecd.ai/app/uploads/2025/05/policy-brief-generative-ai-jobs-and-policy-response-innovation-workshop-montreal-2023-1.pdf
56.
Source: sciencedirect.com
Link:https://www.sciencedirect.com/science/article/pii/S0148296326003486
57.
Source: sciencedirect.com
Title: Air Traffic Controller Trust in Automation in Next Gen
Link:https://www.sciencedirect.com/science/article/pii/S2351978915005107
58.
Source: sciencedirect.com
Link:https://www.sciencedirect.com/science/article/pii/S0003687024002333
59.
Source: internationalaisafetyreport.org
Title: international ai safety report 2026 1
Link:https://internationalaisafetyreport.org/sites/default/files/2026-02/international-ai-safety-report-2026_1.pdf
Source snippet
International AI Safety ReportInternational AI Safety Report 2026...
60.
Source: publications.jrc.ec.europa.eu
Link:https://publications.jrc.ec.europa.eu/repository/handle/JRC147505
61.
Source: ai-act-service-desk.ec.europa.eu
Link:https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-14
62.
Source: consilium.europa.eu
Title: eu Artificial intelligence act
Link:https://www.consilium.europa.eu/en/policies/artificial-intelligence-act/
63.
Source: regulation-ai.eu
Title: article 14
Link:https://www.regulation-ai.eu/en/articles/article-14/
64.
Source: internationalaisafetyreport.org
Link:https://internationalaisafetyreport.org/publication/2026-report-extended-summary-policymakers
65.
Source: internationalaisafetyreport.org
Link:https://internationalaisafetyreport.org/publication/first-key-update-capabilities-and-risk-implications
66.
Source: GOV.UK
Title: international ai safety report 2025
Link:https://www.gov.uk/government/publications/international-ai-safety-report-2025/international-ai-safety-report-2025
67.
Source: policy-lab.ec.europa.eu
Title: eu Understanding the impact of Human-AI interaction on discrimination
Link:https://policy-lab.ec.europa.eu/stories/understanding-impact-human-ai-interaction-discrimination-2025-01-10_en
68.
Source: op.europa.eu
Link:https://op.europa.eu/en/publication-detail/-/publication/68b91f8f-cf0a-11ef-be2a-01aa75ed71a1/language-en
69.
Source: internationalaisafetyreport.org
Title: international ai safety report 2026
Link:https://internationalaisafetyreport.org/sites/default/files/2026-02/international-ai-safety-report-2026.pdf?pubDate=20260710
70.
Source: internationalaisafetyreport.org
Title: international ai safety report 2026 web eng
Link:https://internationalaisafetyreport.org/sites/default/files/2026-02/international-ai-safety-report-2026-web-eng.pdf
71.
Source: internationalaisafetyreport.org
Title: International AI Safety Report
Link:https://internationalaisafetyreport.org/
72.
Source: internationalaisafetyreport.org
Title: INTERNATIONA L AI SAFETY REPORT
Link:https://internationalaisafetyreport.org/sites/default/files/2026-02/ai-safety-report-2026-extended-summary-for-policymakers.pdf
73.
Source: internationalaisafetyreport.org
Link:https://internationalaisafetyreport.org/sites/default/files/2026-02/international-ai-safety-report-2026-executive-summary_1.pdf
74.
Source: internationalaisafetyreport.org
Title: international ai safety report 2025
Link:https://internationalaisafetyreport.org/publication/international-ai-safety-report-2025
75.
Source: internationalaisafetyreport.org
Title: rapport international sur la securite de l ia 2026 resume
Link:https://internationalaisafetyreport.org/sites/default/files/2026-02/rapport-international-sur-la-securite-de-l-ia-2026-resume.pdf
76.
Source: internationalaisafetyreport.org
Title: International AI safety report
Link:https://internationalaisafetyreport.org/sites/default/files/2025-10/international_ai_safety_report_2025_english.pdf
77.
Source: internationalaisafetyreport.org
Title: Resumen ampliado para responsables políticos
Link:https://internationalaisafetyreport.org/sites/default/files/2026-02/resumen-ampliado-para-responsables-politicos-2026.pdf
78.
Source: eur-lex.europa.eu
Link:https://eur-lex.europa.eu/EN/legal-content/summary/rules-for-trustworthy-artificial-intelligence-in-the-eu.html
79.
Source: ai-act-service-desk.ec.europa.eu
Link:https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-26
80.
Source: ai-act-service-desk.ec.europa.eu
Link:https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-6
81.
Source: digital-strategy.ec.europa.eu
Link:https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
82.
Source: digital-strategy.ec.europa.eu
Link:https://digital-strategy.ec.europa.eu/en/policies/ai-act-governance-and-enforcement
83.
Source: rivista.ai
Link:https://www.rivista.ai/wp-content/uploads/2025/02/2501.16946v2.pdf
Additional References
84.
Source: youtube.com
Title: The [Catastrophic]({{ ‘catastrophic-misuse/’ | relative_url }}) Risks of AI — and a Safer Path | Yoshua Bengio | TED
Link:https://www.youtube.com/watch?v=qe9QSCF-d88
Source snippet
Existential Risk and the Future of Humanity: Lessons from AI, Pandemics & Nuclear Threats (Toby Ord)...
85.
Source: youtube.com
Title: The end of human agency
Link:https://www.youtube.com/watch?v=0BjrHB8aNm0
Source snippet
Gradual Human Disempowerment by AI existential risk Gradual disempowerment: Systemic existential risks from incremental AI development |...
86.
Source: youtube.com
Link:https://www.youtube.com/watch?v=bVV91jt56-o
Source snippet
Gradual Disempowerment by Nora Ammann...
87.
Source: youtube.com
Title: Gradual Disempowerment by Nora Ammann
Link:https://www.youtube.com/watch?v=AmtKdoeYGn0
Source snippet
The Catastrophic Risks of AI — and a Safer Path | Yoshua Bengio | TED...
88.
Source: doi.org
Link:https://doi.org/10.1177/154193120605000303
89.
Source: openreview.net
Link:https://openreview.net/pdf?id=mWlnrtOTtm
90.
Source: tandfonline.com
Link:https://www.tandfonline.com/doi/abs/10.1207/s15327108ijap0302_2
91.
Source: rivista.ai
Link:https://www.rivista.ai/wp-content/uploads/2024/05/73d417f9-en.pdf
92.
Source: oecd-ilibrary.org
Link:https://www.oecd-ilibrary.org/en/publications/how-is-ai-changing-the-way-workers-perform-their-jobs-and-the-skills-they-require_8dc62c72-en.html
93.
Source: aiact.annotated.nl
Link:https://aiact.annotated.nl/en/aiact/art/14/