Criminal Law Library Blog
Overview
In his August 26, 2026 essay, “The Turbulent AI Era Is Here. The Choices We Make Now Are Critical,” Bill Gates presents artificial intelligence as a technological transition potentially different in important respects from earlier waves of automation. AI can increasingly perform cognitive work, can spread rapidly through technologies and infrastructure already in widespread use, and can be accessed through ordinary language without requiring users to master specialized computer skills. Gates is not arguing against AI. On the contrary, he remains strongly optimistic about its potential in medicine, education, agriculture, scientific research, government services, clean energy, and assistance for people who otherwise lack access to specialized expertise. His concern is that these benefits will not necessarily be distributed fairly and that significant social and economic disruption could accompany them.
Gates identifies three major categories of concern:
First is the possibility of permanent disruption of employment. Gates believes AI may eventually eliminate significant numbers of both white collar and blue-collar jobs. He is particularly concerned about younger workers because entry level positions the traditional route through which people acquire professional experience may be among those most vulnerable. His overall concern goes beyond unemployment. What happens to an economy and society organized around employment if substantially fewer people are needed to work? Employment provides not only income but, for many people, social relationships, identity, dignity, and a sense of purpose.
Second, Gates warns that AI can increase people’s ability to cause harm. Artificial intelligence can lower the expertise, time, and resources required to commit fraud, produce convincing disinformation and deepfakes, conduct cyberattacks, or engage in sophisticated surveillance. Gates points to possible threats against hospitals, financial institutions, water systems, electrical grids, and government services, as well as biological threats and autonomous weapons. Beyond deliberate misuse, he also acknowledges the longer term possibility that increasingly capable AI systems could behave in ways their designers did not intend or could not adequately control.
Third, Gates is concerned about AI’s effects on children, human relationships, education, and critical thinking. Highly accommodating AI companions could affect the development of genuine interpersonal relationships. Students could also increasingly use AI to avoid the intellectual effort through which genuine learning occurs. Gates therefore emphasizes preserving what educational researchers sometimes describe as “productive struggle”, working through difficult questions rather than simply obtaining an AI-generated answer.
Yet Gates’s essay is ultimately not pessimistic. He calls for serious attention to AI’s dangers while maintaining what he describes as “grounded optimism” about its potential benefits. AI could accelerate medical research, improve diagnosis, expand educational opportunity, assist farmers and communities lacking specialized expertise, simplify access to government services, advance scientific discovery, and give individuals and small organizations access to capabilities that once required substantial financial or institutional resources.
The critical qualification is that although AI can produce these benefits; Gates does not believe market forces alone will necessarily ensure that they are distributed equitably. He therefore proposes several possible responses. These include national and international institutions capable of coordinating the AI transition; consideration of a “Human Reserved” category in which society deliberately preserves certain roles for people even when machines could technically perform them; and changes in taxation so that public policy does not unintentionally encourage businesses to replace workers with machines.
Underlying these proposals is a broader argument: decisions affecting employment, education, security, health, and fundamental social institutions should not be left solely to technology companies. Governments, educators, workers, health professionals, community leaders, technologists, and others should participate in determining how AI is incorporated into society. The central question raised by Gates’s essay is therefore no longer simply How powerful will AI become? Increasingly, it is: How should society organize itself when highly capable artificial intelligence becomes widely available? That question invites comparison with other recent research.
Gates’s Warnings in Perspective: What Other Recent Research Tells Us
Recent research from the International Labour Organization (ILO), Organisation for Economic Co-operation and Development (OECD), International Monetary Fund (IMF), and other organizations supports some of Gates’s concerns while qualifying others. Taken together, these studies suggest that AI could produce substantial benefits and substantial disruption simultaneously, but that important questions about employment, inequality, education, and long-term risk remain unresolved.
Will AI Eliminate Jobs or Transform Them?
Gates’s warning about permanent job displacement is among his most consequential predictions. Recent evidence provides reasons to take that possibility seriously, but it does not establish that mass technological unemployment is inevitable. A June 2026 ILO review found genuine but uneven productivity gains from generative AI while concluding that large-scale employment displacement has so far remained limited. Current evidence points toward transformation of work as much as outright job elimination, although the ILO identifies concerns involving younger workers, inequality, worker autonomy, and job quality. The OECD similarly emphasizes that AI can simultaneously automate existing tasks, create new tasks and occupations, and increase worker productivity.
Particularly important is the distinction between exposure to AI and susceptibility to automation. Lawyers, librarians, researchers, managers, and other professionals may perform many tasks that AI can assist with without their occupations necessarily disappearing. Judgment, accountability, interpersonal communication, contextual understanding, and institutional knowledge may remain important even as individual tasks are automated. There is therefore an important distinction between saying: AI can perform many tasks associated with an occupation and concluding: AI can replace the occupation itself. For many knowledge professionals, the more immediate question may not be whether their occupations disappear, but how those occupations can be reconfigured as particular tasks migrate from humans to machines.
Employment Is Only Part of the Economic Question
IMF research adds another dimension: the significant economic consequence may not simply be unemployment but how AI’s gains and losses are distributed. Workers possessing AI-related skills may receive productivity and wage benefits, while others face displacement or declining opportunities. Younger workers and portions of the middle-skilled labor market may be particularly vulnerable. The question therefore extends beyond Will AI eliminate jobs? It also becomes Who will receive the productivity and income gains produced by AI, and who will bear the costs of the transition?
The same issue exists internationally. The United Nations Conference on Trade and Development (UNCTAD) has warned that AI capabilities, infrastructure, investment, talent, and technological ownership remain concentrated among a relatively small number of countries and companies. This creates another important distinction. AI could democratize access to expertise while simultaneously concentrating economic and technological power. A person or small organization may gain inexpensive access to sophisticated analytical capabilities that were once available only to large institutions. At the same time, the underlying models, computing infrastructure, data resources, and investment necessary to produce those capabilities may remain concentrated. Whether AI ultimately reduces or increases inequality will therefore depend partly upon choices involving education, infrastructure, competition, taxation, access, and governance.
Present Risks and Uncertain Future Risks
Gates also warns about cyberattacks, fraud, deepfakes, biological threats, surveillance, autonomous weapons, and the possibility that future AI systems could behave contrary to human interests.
The International AI Safety Report 2026 provides a useful distinction between risks already producing observable harms and more extreme future risks whose probability remains uncertain.
AI-assisted cyber threats, fraud, manipulation, and other forms of misuse are increasingly concrete concerns. As AI becomes more capable and accessible, it can lower some of the barriers that previously limited sophisticated harmful activities.
More extreme scenarios involving catastrophic loss of human control remain considerably less certain.
This distinction argues against both complacency and inevitability. Some AI risks are already observable; others remain uncertain but potentially severe. Policymakers must therefore decide how much precaution is justified when probabilities remain uncertain but possible consequences could be extraordinary.
Education, Research, and Critical Thinking
Gates’s concern that AI could weaken critical thinking receives qualified support from the OECD’s Digital Education Outlook 2026. An important distinction is between performing better with AI and actually learning more because of AI. Students using general-purpose generative AI may sometimes produce better immediate work without achieving corresponding long-term learning gains. Excessive reliance can encourage cognitive offloading, allowing technology to perform intellectual work students previously had to undertake themselves. But this is not an argument for excluding AI from education.
Properly designed educational AI can potentially improve learning and strengthen critical thinking, creativity, and collaboration. An AI system that simply provides a finished answer may have a very different educational effect from one that asks questions, challenges assumptions, requests evidence, or guides a student through a difficult problem. The critical issue may therefore be less whether students use AI than how they use it.
The same distinction applies to research. AI can accelerate literature review, information processing, brainstorming, comparison, and analysis. But it can also generate plausible inaccuracies, nonexistent references, misattributed citations, and confident conclusions unsupported by reliable evidence. This creates an important paradox: AI may dramatically increase our ability to locate, process, summarize, and generate information while simultaneously making verification of that information more important. For librarians and researchers, source evaluation, provenance, authentication, citation verification, context, and independent judgment may therefore become more, not less, important in an AI-assisted information environment.
Areas of Agreement—and Continuing Uncertainty
Across these perspectives, several points of agreement emerge. AI is likely to change the organization of work; some occupations and workers will be affected more than others; education and adaptation will become increasingly important; and cybersecurity, misinformation, reliability, and other forms of misuse require serious attention. There is considerably less certainty about the magnitude and permanence of future job losses, which occupations will disappear, whether new economic activity will offset displacement, and whether future AI systems will create risks fundamentally different from those associated with today’s technology. Uncertainty should not be mistaken for evidence that the risks are insignificant. Nor should plausible risks be presented as proof that the most disruptive predictions will occur. The evidence instead supports a more measured conclusion: AI is likely to produce substantial benefits and substantial disruption simultaneously, while the distribution of both may be highly uneven.
From Prediction to Choice
Seen in this broader context, perhaps the strongest aspect of Gates’s argument is not any particular prediction about how many jobs AI will eliminate or how quickly particular capabilities will develop. It is his insistence that society should begin addressing these questions before their answers become unavoidable.
The ILO’s emphasis on workers and job quality, the OECD’s focus on skills and education, the IMF’s concern about economic polarization, the International AI Safety Report’s attention to emerging risks, and UNCTAD’s concern about global inequality approach the AI transition from different directions. Yet all point toward a common conclusion: Technological capability alone will not determine AI’s consequences. Public policy, institutional design, education, corporate decisions, professional standards, and individual choices will all matter.
In that respect, Gates’s central warning may be less a prediction than a challenge. Artificial intelligence may eventually become extraordinarily capable, but the social consequences of that capability are not predetermined. The decisions being made now about how AI is developed, deployed, governed, and incorporated into human institutions will help determine whether its benefits are broadly shared, and whether its most serious harms can be contained.
Why Librarians and Researchers Should Care
For librarians and researchers, the rapid development of artificial intelligence presents a paradox. AI can make it dramatically easier to locate, organize, summarize, compare, and generate information. Yet the same capabilities can produce convincing inaccuracies, nonexistent or misattributed citations, synthetic images and documents, and other material that may be difficult to distinguish from reliable information. As AI-generated content becomes increasingly sophisticated, the ability to obtain an answer may become less important than the ability to determine whether the answer deserves to be trusted. Source evaluation, provenance, authentication, citation verification, context, and independent judgment therefore take on renewed importance.
This is also why Gates’s concern about preserving “productive struggle” extends beyond the classroom. Critical thinking develops partly through the process of questioning assumptions, examining competing evidence, identifying weaknesses in arguments, and reaching conclusions that can be explained and defended. If AI routinely performs those functions for us, greater efficiency could come at the cost of diminished intellectual independence.
For librarians and researchers, the challenge is therefore not simply learning how to use increasingly powerful AI tools. It is learning how to use them without surrendering the human judgment necessary to evaluate their output. AI can assist with research and reasoning, sometimes extraordinarily well, but responsibility for determining what is credible, relevant, sufficiently supported, and appropriate to rely upon ultimately remains human. In this sense, the growth of artificial intelligence does not diminish the importance of librarianship and careful research. It reinforces the enduring value of the principles at their core: critical inquiry, verification, intellectual independence, and responsible use of information.
References and Additional Resources
Primary Source
Bill Gates. “The Turbulent AI Era Is Here. The Choices We Make Now Are Critical.” Gates Notes, August 26, 2026. Gates discusses AI’s potential benefits alongside three principal areas of concern—employment disruption, increased capacity for harmful uses, and effects on children, human relationships, education, and critical thinking. He also considers possible institutional, employment, and taxation responses.
Employment, Skills, and Economic Effects
International Labour Organization (ILO). The Impact of GenAI on Jobs, Productivity and Work Organization: A Review of the Empirical Evidence. June 1, 2026. Reviews evidence from experiments, firms, platforms, and worker surveys concerning productivity, employment, workplace organization, younger workers, and job quality.
Organisation for Economic Co-operation and Development (OECD). Skills in the AI Age. OECD Artificial Intelligence Papers, No. 60, 2026. Examines changing skill requirements and the relationships among AI exposure, automation, augmentation, productivity, and employment.
International Monetary Fund (IMF). Florence Jaumotte, Jaden Kim, David Koll, Elmer Li, Longji Li, Giovanni Melina, Alina Song & Marina Mendes Tavares. Bridging Skill Gaps for the Future: New Jobs Creation in the AI Age. IMF Staff Discussion Note 2026/001, January 14, 2026. Examines emerging AI-related skills, employment creation, wage effects, and possible labor-market polarization.
Education and Critical Thinking
Organisation for Economic Co-operation and Development (OECD). OECD Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education. OECD Publishing, January 19, 2026. Examines the distinction between improved performance with AI and genuine learning, along with ways generative AI can be used to support educational objectives.
AI Safety and Emerging Risks
International AI Safety Report. International AI Safety Report 2026. February 3, 2026. Led by Yoshua Bengio and authored by more than 100 AI experts, the report reviews scientific evidence concerning capabilities, emerging risks, and risk-management challenges associated with general-purpose AI.
Global Development and Inequality
United Nations Conference on Trade and Development (UNCTAD). Technology and Innovation Report 2025: Inclusive Artificial Intelligence for Development. April 7, 2025. Examines AI’s potential contribution to economic and social development while warning that unequal access to infrastructure, data, skills, investment, and participation in AI governance could deepen existing global divisions