⚡ Quick Summary
Published by AI Now Institute, this report examines the concentration of power behind the current large-scale AI industry and argues that AI should be understood as a question of power rather than technological progress. It focuses principally on US market and policy developments while also discussing the US–China “AI arms race,” EU policy developments, and investment in other regions. Its account connects dominant firms’ control of compute, cloud infrastructure, data, distribution channels, energy and public-policy agendas to the effects of AI deployment on workers, public services, consumers and communities.
The report contests the narratives supporting industry expansion, including AGI claims, the “bigger-is-better” paradigm, national-security framing, and the recasting of regulation as an obstacle to innovation. It then documents the report’s view of AI’s record in education, healthcare, public benefits, workplaces, housing and immigration enforcement. Its practical deliverable is a roadmap built around five levers: targeting industry harms to everyday people, worker organising, a “zero-trust” policy agenda, stronger networks across expertise and advocacy, and public-centred innovation not organised around large-scale AI.
🧩 What’s Covered
- Executive summary and epilogue: Sets out the report’s central account of “Artificial Power”: AI as both a means through which technology firms consolidate influence and a contingent form of power that can be disrupted. It contrasts the current trajectory with desired conditions for good jobs, shared prosperity, autonomy, sustainability, public services, security, a competitive technology ecosystem and democratic institutions.
- AI’s False Gods: Examines the “AGI mythology,” including uncertainty over definitions, timelines and measurement; the capital, chip, data-centre and energy build-out behind scale-driven AI; the US–China arms-race framing and industrial policy; and arguments that portray regulation as anti-innovation. It also discusses smaller-model rhetoric and the “abundance” agenda.
- Heads I Win, Tails You Lose: Maps mechanisms that, according to the report, protect Big Tech’s advantage regardless of AI demand. These include cloud-provider dependence, partnerships between hyperscalers and model developers, control over enterprise and consumer ecosystems, data-centre construction, utility contracts, energy projections, local subsidies and lobbying against consumer-protection measures.
- Consulting the Record: Draws together examples from education, healthcare, agriculture, public-benefit administration, housing, work and immigration enforcement. It presents five takeaways: benefits are overstated and underproven; AI displaces grounded expertise; solutionism can facilitate austerity; productivity benefits flow to firms rather than workers; and coercive uses undermine rights and due process.
- Technical and evaluative concerns: Describes memorisation and data leakage, jailbreaks and poisoning attacks, hallucinations, discriminatory outputs, emotion-recognition claims, non-peer-reviewed industry research, flawed methods and self-dealing benchmarks. Case discussions include Epic’s sepsis model, MyCity’s chatbot, algorithmic benefit decisions and facial-recognition deployment.
- A Roadmap for Action: Proposes advocacy around DOGE, data centres, algorithmic prices and wages; labour organising across sectors; bright-line prohibitions on listed harmful uses; regulation throughout the AI life cycle; transparency and disclosure duties; independent oversight; and data-focused remedies such as algorithmic deletion.
💡 Why it matters?
The report is relevant to people governing or procuring AI because it shifts attention from a single model’s performance to the institutional dependencies and power relationships surrounding deployment. It identifies concrete settings where mistakes or opacity can affect benefits, employment, housing, healthcare and immigration outcomes, and it links cloud concentration and infrastructure expansion to accountability questions. For policymakers and advocates, its “zero-trust” approach supplies a set of regulatory directions: enforce existing law, use clear prohibitions where safeguards are inadequate, distribute duties across the supply chain, and avoid leaving evaluation solely to vendors.
❓ What’s Missing
The supplied material is only a partial extraction, ending on page 80 during the roadmap’s discussion of independent oversight. It therefore does not permit an assessment of the remainder of Chapter 4, the endnotes, or any concluding material beyond that point. Within the available pages, the roadmap is expressly presented as high-level: it identifies campaign levers, policy directions and examples, but does not provide model legislative text, a universal implementation sequence, or a comparative cost analysis for its proposals. The report’s analysis is also strongly centred on US institutions and policy debates, although it includes selected international developments and describes its agenda as relevant internationally.
👥 Best For
This report is best suited to public-interest advocates, labour organisers, state and local policymakers, public-sector procurement teams, and researchers examining AI market concentration. It is particularly useful for people developing campaigns or policy positions on workplace AI, automated public services, data-centre impacts, surveillance, algorithmic pricing, and independent accountability for AI developers and deployers.
📄 Source Details
Artificial Power 2025 Landscape Report is an English-language report by AI Now Institute, authored by Kate Brennan, Amba Kak, and Dr. Sarah Myers West and dated June 3, 2025. The document gives the citation URL https://ainowinstitute.org/2025-landscape. The supplied text extraction is partial: it covers PDF pages 1–80 of a 123-page PDF, ending during Chapter 4.
