Who Controls Artificial Intelligence? Big Tech, AI Governance, and the Risks of Concentrated Power

Artificial intelligence is changing how people search for information, write documents, create software, analyze data, and communicate. As these systems become more capable, an important question deserves greater attention: who actually controls artificial intelligence?

The answer is more complicated than naming a handful of technology companies. AI involves developers, cloud infrastructure providers, chip manufacturers, businesses deploying AI systems, governments, researchers, and individual users. Each group has different responsibilities and levels of influence.

Nevertheless, concerns about concentrated power have become increasingly visible. In October 2026, actor and United Nations digital governance advocate Joseph Gordon-Levitt attracted attention for questioning whether a small number of companies should have so much influence over the future of AI.

Understanding who controls AI matters because these technologies are increasingly integrated into everyday services and important business decisions.

What Does It Mean to Control Artificial Intelligence?

Control can refer to several different things. One company might own an AI model, another might provide the computing infrastructure required to operate it, and a third might decide how the model is used in a customer-facing product.

For example, a business could purchase access to a commercial AI service, connect it to an internal database, and allow employees to ask questions about company information.

In that arrangement, the model developer determines many of the system’s underlying capabilities. The cloud provider may manage important infrastructure. The business decides what data to connect and which employees can access the tool.

No single participant necessarily controls every part of the system.

This distinction matters when something goes wrong. A misleading answer might originate from model behavior, inaccurate company data, poorly designed instructions, or insufficient human review.

Effective accountability requires understanding which organization is responsible for each part of the workflow.

The Companies Developing Advanced AI Models

Several organizations develop advanced general-purpose AI models. They include companies such as OpenAI, Google, Anthropic, and Meta, alongside other commercial and research organizations around the world.

These developers make important decisions about model architecture, training methods, available features, access policies, and safety measures.

They also decide how models are distributed. Some systems are offered through consumer applications, others through developer APIs or enterprise products, and some model weights are made available under specific licenses.

Different distribution models create different forms of control.

A proprietary service may provide centralized updates and managed security features, but users can become dependent on the provider’s pricing and policies. A model that can be deployed independently may provide greater operational flexibility, although it also places more responsibility on the organization running it.

Why Computing Infrastructure Matters

Training and operating advanced AI systems requires substantial computing resources.

High-performance processors, memory, networking equipment, storage systems, and data centers all contribute to the infrastructure supporting AI.

Organizations with access to large computing budgets may have advantages in training powerful models and serving large numbers of users.

This creates a relationship between AI developers, semiconductor manufacturers, cloud providers and data center operators.

Control over infrastructure does not automatically mean control over every AI application. However, infrastructure availability can influence which organizations are able to compete and how much their services cost.

For businesses, this is one reason to consider the infrastructure and vendor dependencies behind an AI service rather than evaluating only its interface.

Why Are People Concerned About Big Tech’s AI Influence?

Technology markets often reward scale. Companies with large user bases, established distribution channels, and substantial financial resources may be able to expand faster than smaller competitors.

In AI, these advantages can include access to computing capacity, specialized talent, customer relationships, and the ability to integrate AI features into existing products.

Concerns arise when these advantages make it difficult for competing products to gain meaningful access to customers or infrastructure.

For example, if a company relies on one AI provider for search, document processing, and customer service, replacing that provider could require significant changes to workflows and employee training.

This is known as vendor lock-in. It is not unique to artificial intelligence, but the growing number of AI integrations can make the issue more significant.

Competition can encourage innovation, lower prices, and improve services. Concentrated markets may create different incentives. The effects depend on market conditions and should be evaluated using evidence rather than assuming every large technology company behaves identically.

Joseph Gordon-Levitt’s Warning About Corporate Control

In October 2026, Joseph Gordon-Levitt discussed the growing influence of AI companies in an interview with The New York Times.

His argument focused on the people and organizations deciding how AI is developed, deployed, and monetized.

He raised concerns about the possibility that commercial incentives could shape technologies affecting work, communication, and personal relationships without sufficient public oversight.

His perspective reflects a broader debate about the role of private organizations in determining the future of widely used digital infrastructure.

Read our full explanation of Joseph Gordon-Levitt’s AI warning for the context behind his remarks.

What Is AI Governance?

AI governance refers to the rules, processes, and responsibilities used to guide the development and deployment of artificial intelligence.

Governance can operate at several levels. Governments may establish legal requirements. Companies may adopt internal policies. Industry organizations may develop technical standards. Independent researchers may evaluate systems and publish findings.

Within a business, governance can be as practical as deciding which AI applications employees are allowed to use and what information they may share.

A useful governance framework generally addresses questions about responsibility, transparency, privacy, security, fairness, and human oversight.

It should also define what happens when an AI system makes a serious mistake.

Why Transparency Matters

Transparency helps users and organizations understand the capabilities and limitations of AI systems.

That does not necessarily require every company to disclose proprietary source code or confidential technical information.

Useful transparency may include clear descriptions of intended uses, known limitations, privacy practices, data retention, system updates, and available mechanisms for reporting problems.

For example, a company adopting AI-generated customer support should understand whether conversations are stored, how long records are retained, and whether employees can review disputed responses.

Without this information, customers and businesses may struggle to evaluate the risks of using a service.

Can Governments Regulate Artificial Intelligence?

Governments can regulate AI-related activities through privacy laws, consumer protection rules, competition law, sector-specific requirements and dedicated AI legislation.

Different jurisdictions take different approaches, and the requirements applicable to a product may depend on its purpose and where it is offered.

AI used to generate entertainment content does not necessarily create the same risks as AI used in healthcare, employment or financial decisions.

Regulation therefore involves balancing innovation with safeguards appropriate to the potential consequences.

Supporters of stronger oversight argue that companies should not be the sole judges of their own products’ safety and societal impact.

Others caution that poorly designed rules could impose disproportionate costs on smaller developers, limit useful experimentation, or entrench existing large providers.

Both concerns deserve consideration when evaluating specific policies.

Open-Weight AI vs Proprietary AI

One important part of the control debate concerns how models are distributed.

Proprietary AI services generally provide access through interfaces controlled by the developer or service provider.

Open-weight models make trained model parameters available under specified conditions. Depending on their licenses and technical requirements, organizations may be able to deploy or modify them independently.

Open-weight does not automatically mean fully open-source. Training datasets, code, documentation, and licensing rights may differ.

Independent deployment can offer more flexibility over infrastructure and data handling, but it also requires technical expertise, security management, and ongoing maintenance.

For many businesses, the choice is not simply between freedom and restriction. It involves trade-offs among cost, reliability, support, customization, and operational responsibility.

How AI Control Affects Privacy

AI systems frequently process information provided by users, employees, and connected applications.

Depending on the product, this information may include questions, documents, customer records, or other business data.

Privacy depends on what data is collected, how it is processed, who can access it, and which contractual or technical controls apply.

Users should not assume that every AI service stores information in the same way.

Some products offer enterprise controls or different data-handling arrangements. Others may provide consumer settings for managing history and personalization.

Before adopting a tool, review the actual terms and privacy documentation for the relevant product and plan.

For practical guidance, visit our AI Privacy Risks in 2026 guide.

Who Is Responsible When AI Makes a Mistake?

Responsibility depends on the circumstances, applicable law, and the role of each organization.

Consider an AI system used to summarize internal business reports. If it generates an incorrect figure, the organization using the system should have a process for verifying important information before acting on it.

If the underlying service experiences a security incident, the provider’s obligations may be governed by contractual and legal requirements.

In higher-impact applications, responsibility can be more complicated because developers, deployers, and human decision-makers may each influence the outcome.

This is why organizations should define who reviews AI-generated recommendations, who approves important decisions, and how incidents are reported.

What Businesses Should Do Before Adopting AI

Businesses can reduce unnecessary risks by treating AI adoption as a technology management decision rather than simply activating a new software feature.

• Identify the tasks AI will perform.
• Classify the data employees may submit.
• Review the provider’s privacy and security documentation.
• Confirm contractual terms and retention policies.
• Define which outputs require human verification.
• Evaluate alternative providers and data portability.
• Train employees on acceptable use.
• Review performance and incidents periodically.

These steps are useful for organizations of many sizes. They help establish accountability without assuming that every business needs an elaborate governance program.

What Role Do Consumers Have?

Consumers influence technology markets through their choices, feedback, and participation in public discussion.

Individuals can compare privacy policies, evaluate subscription terms, and select products that provide appropriate controls.

They can also support meaningful transparency and participate in public consultations where available.

However, individual choices cannot solve every structural problem. Issues involving market concentration, competition and public accountability may require collective responses from institutions and policymakers.

The Future of AI Control

The organizations controlling AI infrastructure and products may change as technology develops.

New models, more efficient computing methods, alternative deployment approaches, and regulatory developments could influence the competitive landscape.

It is therefore premature to assume that the current distribution of influence will remain unchanged indefinitely.

The enduring question is how to ensure that people and organizations using AI retain meaningful choices and protections.

Final Thoughts

Artificial intelligence is not controlled by a single company or institution. Its development involves a complex network of model developers, infrastructure providers, businesses, governments, and users.

Nevertheless, concentrated resources and commercial influence make governance an important issue.

Understanding the organizations behind AI products can help consumers and businesses evaluate privacy, reliability, accountability, and long-term dependence.

For additional context, read Joseph Gordon-Levitt’s warning about AI companies and our practical AI privacy guide.

Frequently Asked Questions

Who controls artificial intelligence?

Control is distributed among AI developers, infrastructure providers, organizations deploying AI, governments, and users. Their influence depends on the system and how it is used.

Do a few companies control all AI?

No. Several large organizations have significant influence, but universities, smaller companies, open-weight projects, and public institutions also shape AI research and development.

What is AI governance?

AI governance refers to the policies, processes, and responsibilities intended to ensure AI is developed and used appropriately.

Why is AI market concentration a concern?

Concentration can raise questions about competition, pricing, user choice, infrastructure access, and accountability.

Are open-weight AI models safer?

Not automatically. Open-weight models offer different deployment options, but safety depends on their design, implementation, security controls, and intended use.

How can businesses reduce dependence on one AI provider?

Businesses can evaluate alternative services, document integrations, maintain data portability, and avoid unnecessary reliance on proprietary workflows.