Besides holding the power to change rules suddenly, governments control access to a range of valuable resources. Past experience suggests these resources can be allocated in ways that reflect political favoritism rather than competitive allocation.36 As a result, capital, labor, and public funds can be directed toward firms that are politically connected rather than those that create the greatest economic value, weakening institutions in the process.37 A recent example is Indonesia’s Pusat Data Nasional Sementara (Temporary National Data Center) project. Investigators found that a small group of opportunistic officials misused their contracting power to rig the tender process to benefit Lintasarta, a subsidiary of Indosat Ooredoo Hutchison group, in exchange for kickbacks. As a result, the state lost an estimated 140.8 billion rupiah.38 Checks and balances held, however, and the officials were held accountable. The example provides a cautionary tale; if checks and balances fail and transparency is lacking, such conduct may evolve into systemic rent-seeking dynamics. Research from several countries has shown that politically connected firms tend to survive, yet industries with a higher share of such firms exhibit lower growth and productivity.39 Over time, such misallocation reduces market dynamism, crowding out firms that would have generated more value. This process could diminish the broad-based gains AI could have generated and could concentrate the greatest gains in the hands of those who already hold power. Beyond resource allocation, there is also the risk of firms being drawn into the exercise of state power. Many technology companies collect large volumes of personal information; governments increasingly request access to such data.40 Although there may be legitimate security or law enforcement reasons for such requests and companies have safeguards in place, the rising frequency of these requests normalizes this access mechanism—which may be exploited to further the state’s power over its citizens. Ultimately, the safeguards rely on mutual restraint—with companies pushing back on overreach and governments using such access judiciously. This equilibrium can easily go either way when there is a change in corporate policy or government. Although the three outcomes discussed are distinct, they all stem in part from the imbalance of power between corporations (which supply the technology) and states (which depend on these technologies but are responsible for regulating them). Reducing or eliminating this imbalance is crucial to ensuring that countries do not let the opportunities for AI to deliver positive outcomes for their economies slip away. Part 3 of this Report discusses policies to avoid such outcomes. Dynamics between corporations and individuals: Market structure determines who gains from AI Unlike competition between states or contested bargains between governments and firms, the power imbalances between firms and individuals in the AI economy emerge from features of the AI value chain itself, such as large up-front investments, network effects, data feedback loops, and spillover effects that standard market mechanisms cannot address.41 These characteristics lead to market concentration that benefits capital owners more than users and workers. Treating this concentration as a well-understood market failure rather than an inherent characteristic of the AI value chain helps clarify which policy tools can address it. Data advantages lead to durable market power that benefits AI capital owners more than users AI systems improve with use. Each user interaction generates data that can be used to train better models; better models attract more users; more AI’s Political Impact: Reshaping Power Within and Across Countries 211