The Tech Giant's Gambit: Google, Antitrust Law, and Generative AI
This article was originally published on Medium.
On May 21st this year, Google CEO Sundar Pichai announced several significant advancements in AI at the annual developer conference, “Google I/O 2025.” In addition to releasing upgraded versions of key models like Gemini 2.5, Veo 3, and Imagen 4, Google officially declared its deep integration of AI into existing core services, such as introducing AI-powered Google Search and enhancing personalized smart reply features in Gmail.
Since OpenAI launched GPT-3.5 in 2022, large language models (LLMs) have begun to permeate every aspect of our lives. However, as technology rapidly advances, a question might emerge in everyone’s mind: In the foreseeable future, will tech giants like Google leverage their vast capital and market advantages to further monopolize AI development and subtly stifle innovation? And as AI permeates every corner of our lives, will our freedom of information choice and decision-making be subject to the potential control of these companies?
1 The Beginning: Antitrust Lawsuits
Last August, a U.S. District Court ruled that Google’s search business violated the Sherman Antitrust Act, unlawfully maintaining its monopoly in the search market. The court proposed potential remedies including selling the Chrome browser, divesting the Android operating system, and prohibiting exclusive default search agreements with browser developers and telecom operators. Subsequently, in April of this year, a U.S. Federal Court also ruled that Google violated antitrust law in the digital advertising and ad tech markets and ordered it to sell certain ad tech assets.
Notably, U.S. antitrust law offers various remedies, but in both of these judgments, the courts chose the most disruptive means for businesses - structural remedies, specifically business divestiture or asset stripping. The U.S. Department of Justice emphasized that Google is integrating its vast search and advertising data into the training process of its AI models (such as Gemini). They are concerned that such an advantage could allow Google to re-establish a dominant position similar to its search market monopoly in the emerging and profoundly impactful field of AI.
Clearly, these two lawsuits are not just about holding Google accountable for past actions; they are also pre-emptive warnings against potential future market monopolies. The U.S. government seems unwilling to see AI repeat the mistakes of computers, operating systems, and search engines - again being dominated by a single company. They believe maintaining adequate competition in the AI sector is more crucial than ever. These lawsuits are undoubtedly a clarion call for the future order of the AI industry.
2 Google’s Core Competencies in the GenAI Era
Google’s ability to establish a dominant position in the AI field is not unfounded. Its strength is built upon two decades of deep engagement in the digital world and the inherently high barriers to entry for large language models themselves.
First, Google controls the world’s largest information gateway and a vast digital ecosystem. This omnipresent connection allows Google to seamlessly integrate its AI products into every corner of users’ daily lives. Especially in the search market, Google has held a near twenty-year monopoly, meaning it can direct billions of users to its platform through default search engine agreements. When AI is deeply integrated into search functions, this default effect will seamlessly transfer to the AI realm, making it difficult for other competitors to gain sufficient user reach, even if they possess innovative AI technology.
Second, LLM training itself creates extremely high capital and resource barriers. This is a game only tech giants can play:
- Massive Computational Power Requirements: Training a top-tier LLM model consumes enormous computational resources and electricity costs. This means only a few capital-rich tech giants or cloud service providers with large data centers can afford it.
- Extremely Large Data Volumes: Data is the crucial fuel for LLM success. Google has accumulated the world’s largest, most real-time, and most diverse collection of user behavior data and web content indexes. These are irreplaceable golden resources for training AI models. Google’s search capabilities and comprehensive information ownership offer a unique advantage in retrieval-augmented generation (RAG), forming an extremely high moat.
- Top-Tier AI Technology and Talent: Globally, top AI researchers and engineers with core LLM R&D capabilities are scarce resources. Large tech companies attract these top talents through high salaries and leading research environments.
These three elements collectively create extremely high barriers to entry, making it highly unlikely for the foundational model market for LLMs to exhibit traditional perfect competition. The market will likely end up in an oligopoly, dominated by a few well-capitalized players.
3 The Future Under AI Monopoly: Information, Bias, Content, and Industry Ecosystem
If Google truly monopolizes the AI market, its impact will be even more profound than past search engine monopolies because the nature of AI changes how we acquire information and interact with the world.
3.1 Monopolistic Differences in Information Acquisition: A Deeper “Black Box”
A traditional search engine’s core function is indexing and linking. Users can theoretically still click and visit original websites to independently obtain more detailed, diverse, and unaggregated information.
However, generative AI’s core function is aggregation and generation. Taking Google’s AI Overviews as an example, it no longer just provides links but directly extracts information from multiple sources, then synthesizes and summarizes it, and finally re-generates a concise and seemingly authoritative “final answer.”
This shift in information presentation leads to a deeper “black box” problem. When AI directly provides an answer, it’s difficult for users to know exactly which original websites the answer was extracted from, or whether the model might have introduced “hallucinations” (i.e., incorrect or fabricated information). This significantly reduces information transparency.
3.2 Discriminatory AI and Data Bias
Even if Google has no malicious or discriminatory intent, its AI models may produce discriminatory outputs due to inherent biases in the training data, a phenomenon known as AI bias. This data reflects various inequalities and injustices that have long existed in human society. When a model learns from such biased data, it will inevitably “internalize” and reproduce these biases in its judgments and generations.
These seemingly “objective” algorithmic outputs can, in essence, subtly exacerbate existing social biases and discrimination. When a single company dominates the AI market, any bias in its models can be amplified. This spread of implicit bias is far more difficult to detect and correct because it’s cloaked in the guise of “neutral” technology.
To avoid this situation, we need more diverse, independent LLM models to compete and counterbalance each other. An information environment that reflects diverse perspectives is the foundation of a healthy and resilient society.
3.3 Generative AI’s Disruption to Content Creators
In the search engine era, there was a symbiotic relationship between search engines and content creators. Websites needed Google to bring traffic, and Google needed website content to enrich its search results.
However, generative AI’s “direct answer” characteristic completely breaks this symbiotic relationship. If AI Overviews can directly satisfy users’ information needs without requiring them to click through to original websites, this will directly “hijack” traffic that would otherwise go to content creators. For digital publishers reliant on traffic, ad revenue, or subscription models, this is undoubtedly pulling the rug out from under them.
In the long run, this traffic hijacking could lead to a significant decline in the incentive to produce high-quality original content. The healthy development of the entire digital content ecosystem will be severely impacted, ultimately leading to information impoverishment.
3.4 LLM as Infrastructure for All Industries and AI Agents
In recent years, the concept of AI agent has swept the tech world: past AI was limited to answering questions, but now it’s evolving towards performing tasks. An AI agent is an intelligent agent with the ability to think, plan, and execute. This is a commercial reality that will emerge in the coming years.
When LLMs gain the ability to take action, their role will shift from mere information processors to the underlying operating system for various industries.
In other words, LLMs will become the core engine driving digital transformation and efficiency improvements across all industries. If Google monopolizes LLMs, it means it will control the underlying technology, tools, and data flow for almost all these new services and traditional industry transformations. This will give it enormous leverage and pricing power, potentially leading to a monopoly over nearly all industries.
4 Striking a Balance Between Competition and Innovation
Given that the LLM market is inherently an oligopoly with high barriers to entry, what is the true significance of antitrust law imposing penalties on Google? The goal of antitrust agencies is to combat Google’s “anti-competitive” behaviors that leverage its existing market dominance, thereby promoting limited but effective competition.
4.1 The Core Purpose of Antitrust Action: Promoting “Effective Competition”
Antitrust law focuses on Google’s “abuse” of its data or platform advantage to exclude competition. For example, Google is accused of paying huge sums to mobile manufacturers or browsers to make Google Search the default search engine. This behavior, securing market share through financial power rather than product competitiveness, is considered abuse. This model is highly likely to be replicated for AI assistants in the future.
Furthermore, Google’s vertical integration in the ad tech market has also been accused of creating conflicts of interest and limiting competition. The U.S. Department of Justice believes Google also abuses its dominant position through self-preferencing its services. These are artificially constructed barriers to competition.
Therefore, antitrust strategies should focus on prohibiting these anti-competitive behaviors. The government should explicitly prohibit Google and any other tech giant from using its dominant position in one market to demand exclusive agreements, ensuring that all players can compete fairly on technology and innovation.
4.2 Data Portability and Market Fairness
While directly penalizing Google for possessing vast data is not reasonable, its scale can indeed create extremely high barriers to entry. A more appropriate approach is to expand data protection regulations like GDPR, making them applicable to a broader range of general data, and requiring companies to facilitate data portability when appropriate. This would allow users to easily export their data from Google services and transfer it to other competing services, thereby reducing the cost for users to switch platforms and weakening the lock-in effect of data.
Moreover, in extreme cases, if certain anonymous, aggregated key data is proven to be an “essential facility” for market competition, and its owner abuses its dominant position, then data interoperability should also be considered. This would allow competitors to access data crucial for AI training, thereby helping them improve their own AI models. This is about promoting data flow, thereby weakening the exclusivity and barrier effect of data.
The ultimate goal of these measures is to ensure that even giants like Google, with their immense advantages, must compete in a relatively fair environment, winning the market through product and technological innovation, rather than solely relying on its existing privileges and default advantages. The market can still maintain a certain degree of vitality and diversity, creating space for more diverse AI applications and innovation.
This antitrust battle concerning Google and AI is fundamentally about shaping the future order of the digital economy. It is not merely a legal game but a significant test of whether we can maintain fair market competition and user choice while ensuring technological progress.
I believe our goal should be to promote healthy competition rather than pursuing radical dismemberment. Therefore, I do not support structural remedies like mandatory divestiture of the Chrome browser or Android operating system. A more effective approach should be to egulate clear abuses, such as prohibiting exclusive agreements, restricting self-preferencing, and reducing data barriers by promoting data portability and interoperability, allowing all players to compete fairly on technology and services.
As participants in the digital age, our attention and choices will collectively shape the future of AI - or perhaps, humanity itself.