Unauthorized use of news for AI development threatens journalism

Home > Opinion > Columns

Unauthorized use of news for AI development threatens journalism

Published: 26 Mar. 2025, 00:02 Updated: 26 Mar. 2025, 17:39

Audio report: written by reporters, read by AI

Nam Jeong-ho

The author is the head of the media division of Korea Press Foundation.

Amid the political chaos following the Dec. 3 martial law controversy, a critical issue has been overshadowed — the mounting copyright debate ignited by the “DeepSeek shock” from China. The Chinese artificial intelligence startup DeepSeek stunned industry insiders in Korea by developing an AI model on par with ChatGPT at less than one-twentieth of the cost incurred by its American competitors. While some in the domestic AI industry were taken aback, others responded with excitement, asking, “If China can do it, why can’t we?” This sentiment has fueled a nationwide push — led by the government — to develop a world-class Korean AI model.

OpenAI founder Sam Altman has long advocated the “Scaling Law,” which argues that the performance of AI depends on the size of its parameters, computing power, and — most crucially — the volume of training data. DeepSeek’s success is attributed to its access to an enormous trove of training data at an absurdly low cost. In China, where information is tightly controlled by the state and major outlets like Xinhua and the People’s Daily are state-owned, it is widely assumed that government-backed AI companies had free rein to use this data.

Inspired by China’s approach, voices in Korea are now calling for similar privileges for local developers. In February, the Presidential Committee on the Fourth Industrial Revolution even suggested a radical measure: allowing Korean AI firms unrestricted access to domestic data, with copyright fees to be settled later. This urgency seems driven by fears that Korea may fall behind in the global AI race if it fails to act swiftly. But is it justifiable to sacrifice journalism — one of the nation’s foundational pillars — in pursuit of short-term goals?

The OpenAI logo is seen on a mobile phone in front of a computer screen displaying output from ChatGPT, on March 21, 2023, in Boston. [AP/YONHAP]

Since the emergence of ChatGPT, tensions between media organizations and Big Tech over the use of news content have only intensified. Generative AI systems must continually ingest a wide array of data to function effectively — and among the most valuable data sets are journalistic works. News content, with its rigorous fact-checking, accurate information, and clean grammar, stands out as ideal training material for AI models. Unsurprisingly, developers prefer it. The problem is that many are trying to use this content for free — disregarding the labor and intellectual investment of journalists, and in clear violation of copyright law.

The legal battle began in earnest when The New York Times filed a lawsuit against OpenAI in 2023. Since then, copyright disputes have erupted globally. Recently, major Korean broadcasters — KBS, MBC, and SBS — as well as the Korea News Association have either filed or are preparing to file lawsuits against Naver for similar violations.

The U.S. Supreme Court has yet to issue a landmark ruling on the matter, but an important shift is underway: global AI companies are increasingly resolving copyright issues not through litigation but through compensation agreements with media outlets. As of late February, OpenAI had signed content licensing deals with 38 media organizations, including News Corp and the Financial Times. This trend reflects a growing recognition that content providers are essential partners in building sustainable AI ecosystems. In stark contrast, Korean AI companies remain reluctant to offer compensation — some even paint copyright advocates as self-interested obstructionists standing in the way of progress.

Consider the case of Toyota. In 2009, the automaker teetered on the brink of collapse after a series of unintended acceleration incidents forced the recall of 10 million vehicles. The root cause? Substandard components — the result of excessive cost-cutting. It was a painful lesson in the importance of fair value.

Reporters asking questions to Bu Ji-seok of Law Firm Buyou, center, in front of the Seoul Metropolitan Police Agency in central Seoul on March 17, before reporting YouTuber Lee Jin-ho for defamation to the police. [NEWS1]

In the world of generative AI, news organizations resemble the core parts suppliers in the auto industry. Without timely and reliable content updates, AI models lose their relevance and utility. If media outlets — the lifeblood of quality content — are weakened, AI systems will be left with nothing but inferior data.

Unlike China, Korea’s media landscape is not dominated by state-run outlets shielded from financial risk. If AI becomes a primary distributor of news content, already struggling Korean newsrooms could face even greater hardship. Ensuring that AI companies pay fair value for news content is not charity — it is a path to mutual prosperity. No matter how urgent the moment, we must not make the fatal error of killing the goose that lays the golden eggs.

Translated using generative AI and edited by Korea JoongAng Daily staff.