Japan Plans Antitrust Probe Into AI Search Engines Over Use of News Articles

Google, Microsoft and Perplexity face scrutiny over news use

Japan's competition watchdog plans to investigate whether search engines using generative artificial intelligence are treating news publishers unfairly by using their articles without permission or adequate payment. The Japan Fair Trade Commission (JFTC) is examining practices that could violate the country's Antimonopoly Law, including abuse of a superior bargaining position.

As of January 3, 2026, the planned fact finding investigation targets major providers including Google LLC, Microsoft Corp., Perplexity AI Inc. and Japan's LY Corp. It may also cover other companies, including OpenAI, which operates ChatGPT.

The central concern is that AI search services can turn journalism into direct answers, giving readers information without requiring them to visit the websites that produced it. Publishers could lose advertising income and opportunities to attract subscribers, while technology companies gain value from their reporting.

The inquiry follows a JFTC report published in 2023 that warned about platforms using news articles for nothing or for very low fees. The commission also intends to examine what improvements companies have made on the problems identified in that report.

Japan's action brings competition law into a dispute often framed around copyright. A company may have permission under copyright rules to use material for some kinds of AI training, yet still face scrutiny over the terms it imposes on publishers that depend on its services.

The investigation is at the information gathering stage. Its completion date has not been given, and possible changes to company practices will depend on what the commission finds.

How AI search changes the journey to a news article

Traditional search engines generally respond to a query with a list of links. A reader chooses a result, visits a website and reads the article there. AI search engines can instead accept a question written in ordinary conversational language and produce a combined answer drawing on information from multiple sources.

Generative AI is software that creates text or other material from patterns learned in data. The search services use large language models, systems trained on enormous collections of text, to identify relationships between words and ideas and produce responses resembling human writing.

ChatGPT and Perplexity use systems that ingest internet material, including news articles, books and other publications. Their ability to summarize and combine information allows users to ask complex questions without opening and comparing several websites themselves.

That convenience explains much of the appeal. A reader can receive a direct answer instead of searching through individual pages. Publishers and other creators argue that the same process repurposes their work without proper credit or compensation.

Training and answering a search query are different stages of the technology. Training teaches a model patterns from a collection of material; the resulting service uses its capabilities to respond to users. Both stages matter to the dispute because publishers are concerned about how their work enters AI systems and how those systems subsequently present information.

Why fewer clicks can mean less money for journalism

News websites earn much of their revenue by displaying advertisements to people who visit their pages. If an AI answer satisfies a reader before that visit happens, the publisher loses an opportunity to earn advertising income from the reporting it paid to produce.

Publishers also lose more than a single advertising impression. A visit can introduce a reader to other articles, encourage a subscription and help establish a continuing relationship with the publication. AI summaries can interrupt that process by keeping the reader inside the search service.

Major technology companies already have substantial power in search and digital advertising. Critics argue that incorporating news into AI services without negotiating licenses lets those companies take value from journalism while returning little to the organizations that create it.

The concern extends to the resources available for future reporting. Investigative journalism, local coverage and other work requiring sustained investment become harder to finance when publishers lose income and access to readers. The original reporting that helps inform AI answers still has to be produced by someone.

Licensing income could help address these losses, but publishers also face a broader change in how people find and consume news. Payment for access to articles and the loss of direct audience relationships are connected business problems, each with its own consequences.

The platform disputes that came before AI

News publishers have experienced changes in distribution before. During the 2010s, social media platforms such as Facebook initially sent substantial traffic to news websites. Later changes to algorithms, the rules deciding which posts people see, gave priority to other types of content and reduced referrals to publishers.

News organizations that had arranged their operations around social media distribution suffered when those platforms changed direction. Their experience showed how decisions made by a technology company could affect income at publications outside its business.

Google's role in search has created similar tensions. Publishers have long complained that snippets of articles displayed in search results can answer a reader's question before the reader clicks through to the full story.

AI answers extend that concern because they can combine information into a more complete response. The issue for publishers is therefore both the amount of news being reused and the increasing ability of a search service to satisfy readers without sending them elsewhere.

What Japan's competition law could address

The JFTC's 2023 investigation into news media transactions identified possible problems when major platforms use articles without paying, or pay fees that are substantially too low. It warned that a stronger platform could violate the Antimonopoly Law by using its position to disadvantage a publisher.

The JFTC, Japan's competition regulator, described such conduct in its report as:

an abuse of a superior bargaining position under the Anti-Monopoly Law.

Examples included changing contract terms unilaterally and setting excessively low fees for using content. The planned AI inquiry will revisit those concerns and assess whether companies have made improvements.

A superior bargaining position means one business has enough strength in a trading relationship to impose terms that the other has difficulty refusing. For a publisher dependent on a platform for visibility or readers, rejecting unfavorable conditions can carry a business cost.

The commission believes some uses of news without permission or adequate compensation could fall within this prohibition or other provisions of the law. The practical focus is on the relationship between the companies: who sets the conditions, how content is used and whether the publisher is placed at an unfair disadvantage.

Why Japan's copyright rules add another layer

Japan has historically taken a relatively permissive approach to copyright and AI. A 2019 revision to the Copyright Act included Article 30-4, which allows broad use of copyrighted works for information analysis, including the training of AI models.

The provision allows certain commercial uses as well, helping make Japan comparatively welcoming to developers seeking to analyze copyrighted material. Information analysis means examining data for patterns rather than using a work primarily to experience its writing, images or other expression.

Japan's Agency for Cultural Affairs added detail through draft guidance on AI and copyright released in January 2024. The draft was revised after nearly 25,000 public comments, reflecting substantial public interest in where the boundaries should lie.

The guidance distinguishes analysis from using material for the enjoyment of the thoughts and sentiments expressed in a work. Pure analysis may be permitted without consent. Uses intended to produce outputs experienced as creative expression from copyrighted works, including attempts to mimic particular creators' styles, may require permission.

The guidance also addresses unjust harm to copyright holders. This can arise when AI output competes with a rights holder's market or interferes with potential channels for selling or distributing the work. Publishers' ability to earn money from their articles is therefore relevant to the legal discussion.

Copyright and competition rules address different aspects of the same activity. Copyright concerns the permitted use of a work; competition law concerns business conduct and the exercise of market or bargaining power. An AI company operating in Japan may need to consider both when deciding how to obtain and present news content.

How the dispute developed

Japan's planned inquiry follows several stages of legal and regulatory attention to the relationship between AI developers, platforms and content owners. The key developments described by January 3, 2026, are:

• In the 2010s, publishers gained traffic from social media before changes in platform algorithms reduced news referrals.
• In 2019, Japan revised its Copyright Act to include broad permission under Article 30-4 for information analysis, including AI training.
• In 2023, the JFTC published its investigation into news media transactions, warning about unfair platform terms and inadequate usage fees.
• In January 2024, the Agency for Cultural Affairs released draft AI and copyright guidance, later revised following nearly 25,000 public comments.
• By January 3, 2026, the JFTC had announced plans to investigate AI search engines and examine progress on the issues raised in 2023.

The sequence shows why the new inquiry reaches beyond the mechanics of generating text. Japan is considering how existing copyright permissions operate alongside rules intended to protect businesses from unfair treatment by stronger trading partners.

Europe and the United States take different routes

European authorities are also examining AI search and the power of major platforms. The European Commission has begun investigating Google over alleged violations of European Union competition law.

The EU's Digital Markets Act designates certain large online platforms as gatekeepers and subjects them to specific obligations. These include requirements supporting fair competition and restrictions on giving preferential treatment to their own services. The EU has also adopted the AI Act as part of its wider regulatory framework.

Those rules could address some concerns surrounding AI search, although their application to generative AI is still being developed. The European approach combines competition enforcement with legislation governing large platforms and AI.

The United States has taken a more fragmented approach. Under the Biden administration, the Federal Trade Commission signaled an aggressive use of its existing powers to regulate AI. It warned that discriminatory effects from AI tools and unsupported claims about what AI can do may violate the FTC Act.

The FTC has also acted against companies whose AI technologies deceived or harmed consumers. The United States lacks comprehensive federal legislation specifically governing AI development or deployment, while individual states have adopted their own rules and Congress has considered numerous AI bills.

For businesses, this creates a mixture of state requirements and older federal laws applied to new technology. Japan's inquiry similarly uses existing law, but its immediate focus is the relationship between AI search providers and news publishers.

Publishers are taking AI companies to court

The New York Times has sued OpenAI and Microsoft, alleging copyright infringement involving its news content. The case is part of a broader challenge over whether training AI models on copyrighted works qualifies as fair use under United States copyright law.

Fair use is a legal principle that permits some uses of copyrighted material without permission. The dispute concerns how that principle applies when large quantities of published work are used to develop systems that generate new responses.

Artists, photographers and writers have also filed lawsuits alleging that their work was used to train AI image and text generators. They are seeking compensation for past use and legal precedents that would protect their intellectual property as AI develops.

Perplexity faces several specific challenges. The Chicago Tribune has sued the company, alleging that it used copyrighted articles without permission and bypassed paywalls to generate content. Encyclopedia Britannica and Merriam-Webster have also brought lawsuits alleging copyright and trademark violations.

These cases address claims about particular uses of protected material, while the JFTC is examining business practices under competition law. Court decisions could reshape the legal requirements for developing AI, alongside the effects of regulatory investigations.

Licensing deals offer one business model

OpenAI has reached agreements with the Associated Press and Axel Springer, among other publishers, to obtain access to archived articles for training large language models. The arrangements involve licensing fees and experimental use of OpenAI software.

Such partnerships provide a possible way to pay publishers while giving AI developers agreed access to content. They also show that companies are already trying commercial arrangements alongside the legal disputes over unauthorized use.

The scale remains difficult. AI models are trained on enormous datasets containing billions of articles and other items. Negotiating agreements with potentially millions of rights holders would create substantial administrative work and costs that developers may regard as prohibitive.

Some AI companies therefore argue for broad fair use exemptions. Others have developed mechanisms allowing publishers to opt out of having their content used in training. An opt out gives a content owner a way to request exclusion rather than requiring a separate negotiation over every item.

If the JFTC's findings lead to requirements for permission or compensation, developers may need to change their licensing arrangements, their treatment of news in answers or both. Establishing and operating those arrangements could increase the cost of AI systems.

Publishers face their own negotiation problem. Large technology companies have far greater resources and bargaining strength. Even when regulatory pressure brings companies into talks, publications may struggle to secure favorable terms without collective bargaining power or rules setting standards for fair compensation.

What the investigation could produce

The JFTC's fact finding process is intended to collect information from companies, examine market conditions and assess whether practices may breach existing law. The commission is gathering evidence before deciding what response is appropriate.

Following the investigation, the commission is expected to identify potential Antimonopoly Law violations and recommend improvements. Possible responses include administrative guidance, voluntary commitments from companies and formal enforcement if violations are found.

Similar investigations elsewhere have taken months or years. The immediate task in Japan includes checking whether the concerns identified in the 2023 report have been addressed and whether AI search has introduced further problems in the use of news content.

For technology companies, scrutiny creates an immediate need to examine compliance and a possibility of broader changes to business practices. The investigation itself may encourage adjustments intended to reduce legal risks.

Japan's decisions could also affect discussions in other countries. As a major economy with substantial technological capabilities, it has influence beyond its domestic market. Its use of existing antitrust law offers an approach that other jurisdictions facing comparable disputes may consider.

Finding a workable relationship between AI and news

Publishers hope regulatory action can produce fairer terms and new income from licensing. Some are pursuing lawsuits at the same time, seeking compensation for earlier unauthorized use rather than waiting for regulators to act.

Licensing alone may not replace the advertising revenue and audience relationships disrupted by AI answers. News organizations are exploring subscriptions, memberships, events, consulting and other sources of income, while trying to build more direct relationships with readers and demonstrate the value of professional reporting.

Technology may provide part of the response. Developers are exploring better attribution and ways to send readers to original content. Standards for content provenance, meaning records of where material came from, could help identify creators and support credit and compensation. AI companies are also considering business models that share revenue with content producers.

The wider debate concerns the power of platforms such as Google, Meta and Amazon in the digital economy. Policymakers are applying competition, copyright and consumer protection rules developed before today's generative AI systems, while considering whether those rules need updating.

AI advocates argue that overly restrictive rules could slow technological progress and useful applications. Publishers and creators argue that insufficient protection could weaken their businesses and reduce the quality and diversity of material available online.

AI services depend on good information, and journalism depends on income to keep producing it. Japan's investigation focuses on whether the terms connecting those activities are fair, while licensing, technical changes and new business models offer additional ways to support that relationship.

The Bottom Line

• Japan's JFTC plans to investigate AI search providers including Google, Microsoft, Perplexity and LY Corp.; OpenAI may also be covered.
• The inquiry concerns possible Antimonopoly Law violations involving news use, payment and unfair bargaining terms.
• A 2023 JFTC report warned about platforms paying nothing or excessively low fees and changing terms unilaterally.
• Japan permits broad information analysis under its copyright rules, with limits addressed in subsequent cultural affairs guidance.
• AI answers can reduce visits to news websites, affecting advertising income, subscriptions and reader relationships.
• European scrutiny and lawsuits in the United States are examining related competition and copyright issues.
• Licensing partnerships already exist, but negotiating across vast datasets presents costs and practical difficulties.
• The Japanese investigation is gathering facts, with possible guidance, voluntary commitments or enforcement to follow.