The defining competitions between global powers have rarely been limited to physical assets. During the Cold War, influence was exercised not only through military strength but also through control over information. Today, a new kind of arms race is emerging. Competition isn’t confined to ideological messaging or broadcast media; it’s defined by artificial intelligence systems that are increasingly influencing how information is accessed and interpreted.

Today’s information environment is shaped by systems that operate at unprecedented scale and speed. AI tools are now integrated into search engines, social media, and everyday workflows, filtering and generating information in real time. AI’s control has shifted from managing content generation to shaping how content is presented to individuals.

This shift has significant implications not only for states and institutions, but for how individuals engage with information. The emerging AI arms race isn’t solely about developing more advanced models—it’s about determining who controls the systems that shape information visibility and interpretation. Understanding this transition is key to assessing how information will be accessed, governed, and presented in the future.

From Cold War propaganda to algorithmic influence

During the Cold War, the conflicting regions strictly controlled media and information, framing content to support their interests.

Now, information systems are decentralized—anyone can publish, but visibility is mediated and decided by algorithms and AI. Algorithms determine what content appears, while AI systems filter and summarize it for users.

Control has shifted from producing information to filtering it. Information distribution is now global, real-time, and personalized, with algorithms tailoring content to individual users.

What the AI arms race involves

The AI arms race is often framed in terms of technological development, but it actually involves competition across multiple interconnected layers that determine how AI systems are built, deployed, and accessed.

Compute infrastructure is essential for AI development and deployment. As AI grows, so does the need for infrastructure that scales computing power, like data centers and AI chips. Demand for AI-ready center capacity is estimated to increase at an annual average rate of 33% between 2023 and 2030. Access to compute infrastructure determines who can build and scale AI systems in the first place.

AI also depends on large datasets. Data enables AI systems to learn how to recognize patterns and generate outputs. Data quality and access directly impact model performance. Access to broader or higher-quality data can develop more capable systems, creating an advantage for those with access to extensive digital ecosystems and data sources.

At the core of AI systems are models that are trained to generate responses. Training a model shapes its outputs. Model design and training influence how it interprets prompts and how it presents information. As a result, different models may produce different outputs even when responding to the same prompt.

Beyond development, distribution is crucial in determining how AI systems are used. Users increasingly interact with AI through platforms. Control over these channels determines which systems users rely on. As AI becomes more integrated into digital platforms, distribution shapes both information visibility and the adoption of AI systems at scale.

The competition isn’t just about AI itself—real control comes from combining these four layers.

AI as the new gatekeeper of information

Before, users navigated multiple sources through search engines. Now, AI assistants provide summaries synthesized based on available data.

This shift has changed how people access information—according to DataReportal, over 1 billion people use GenAI platforms and LLMs every month. Many users rely on AI to summarize, filter, and generate information for them. As a result, people have less direct interaction with raw sources.

On platforms like X, users increasingly rely on built-in AI tools like Grok to provide context directly within conversations. Instead of searching independently, users receive immediate, synthesized summaries.

How this impacts human–AI interaction

Users increasingly rely on AI-generated summaries and often accept outputs without manually verifying the accuracy of the information presented. However, AI tools are prone to generating misleading outputs, with inaccurate or even biased content.

The AI may also generate different outputs per user. AI systems are trained on user data, so they can tailor content for each user. As responses become more personalized, users may find different interpretations of the same topic, contributing to increasingly fragmented information environments.

As AI influences how information is accessed and interpreted, it’s also subtly shaping how people interpret information and make decisions.

Diverging global approaches to control

There are three models that influence how information is accessed through AI: state-driven systems, market-driven ecosystems, and regulated hybrid systems.

State-driven AI systems are influenced by the government and control outputs to avoid politically sensitive topics and align with national policies. Governments influence training data and outputs, shaping how sensitive topics are handled.

Market-driven ecosystems are developed by companies like OpenAI and Google. Company-led AI systems are deployed globally, and the nature of their generated responses is regulated by safety policies and usability.

Hybrid systems combine private innovation with regulatory oversight. The European Union’s AI Act sets regulations to promote transparency, accountability, and the ethical use of AI. Outputs generated by hybrid systems may limit certain types of responses or provide disclaimers.

Different AI systems may produce different answers for the same question because they’re shaped by underlying structures.

Infrastructure still determines power

Despite its digital nature, AI remains fundamentally dependent on physical infrastructure. Deploying advanced AI systems requires resource-intensive data centers, a reliable energy supply, and access to specialized hardware. An IEA report revealed that the US had the largest share of global data center electricity consumption in 2024, followed by China and Europe.

However, AI compute infrastructure isn’t evenly distributed. Some regions have the capacity to invest in the required infrastructure, but many lack the resources to develop advanced AI ecosystems or the infrastructure required to compete at scale. This creates an imbalance—some countries shape AI systems while others primarily consume them.

As AI adoption expands, infrastructure development is becoming a strategic asset for governments seeking more control over their local digital ecosystems. Building local infrastructure requires significant investment and coordination between the public and private sectors. In many emerging markets, this process is already underway, with increasing focus on establishing the foundations needed to support AI deployment at scale. Partnerships that align infrastructure development with national priorities will be fundamental in shaping how AI systems are accessed and governed.

Risks and trade-offs

A small number of regions currently control a large portion of AI capability. Building infrastructure is expensive, so only a few players can compete. The downside is that there are fewer systems shaping information, with less diversity in outputs.

AI generates content using complex processes. Users only get outputs without seeing how they’re generated. This limits transparency in information presented and makes it harder to evaluate content accuracy.

AI systems also introduce more subtle forms of influence through the way they frame information. Rather than censoring information outright, these systems often shape understanding by summarizing content, selecting key points, and presenting information in a particular context. Even small differences in emphasis or wording can affect how users interpret a topic. With users increasingly relying on synthesized responses, this form of influence is subtly shaping perception without requiring greater control over information access.

Two Possible Outcomes

Broadly, two possible trajectories are emerging: one defined by convergence around a small number of dominant systems, and another defined by fragmentation across multiple competing ecosystems.

In a convergent outcome, a small number of AI platforms scale globally to dominate information access. Convergence may lead to more standardized experiences across regions, where users rely on similar systems regardless of location. This may improve AI efficiency and accessibility, but it could also concentrate influence over how information is filtered and presented.

Alternatively, a fragmented scenario would have distinct AI ecosystems developing across regions. Differences in regulations and priorities could lead to systems operating under varying constraints and design principles. Users may receive different responses across regions. This can increase diversity in information presentation, but it can also introduce inconsistencies in information access and interpretation.

The outcome that emerges will depend on how these systems are developed, governed, and integrated across different regions.

Conclusion

The influence of AI systems shows a broader shift in how information is accessed and understood. Rather than simply making knowledge more accessible, these systems are increasingly shaping how that information is filtered, interpreted, and presented to users.

As these systems become more integrated across digital environments, their influence goes beyond technical capability and affects how individuals form opinions and make decisions. The structures behind AI, including infrastructure and governance, determine what information is visible and how it’s framed.

The long-term impact will depend not only on technological advances but also on how AI systems are developed, deployed, and governed at scale. As the global landscape continues to evolve, the question isn’t only who builds these systems, but how they shape the information environments that people rely on.