AI Boom or Copyright Doom? Lessons from Asia

Around the world, authors, musicians, and visual artists fear artificial intelligence will replace them, while at the same time, the AI industry exploits their work to train the technology. Not in Japan or Singapore. Both Asian countries have approved AI-friendly copyright laws, convinced that AI can help increase economic and labor productivity in the face of demographic decline.

The Japanese and Singaporean reforms allow copyrighted works to be used for AI text and data mining. They avoid prolonged US-style legal battles between the AI industry and rightsholders about what constitutes “fair use” and the compliance burden of the European Union’s AI Act. At the same time, mounting opposition from copyright holders is nudging both Asian governments to encourage licensing deals.

In 2019, Japan amended its copyright act to include Section 30-4, allowing exploitation of a copyrighted work for “data analysis.” Singapore’s 2021 Section 244 permits use “for the purposes of computational data analysis.” Both exceptions to copyright infringement for data analysis allow AI companies to use copyrighted works without remuneration.

They aim to boost AI investment. A Japanese government report revealed the copyright reform came to meet industry “expectations for new innovations that generate added value by combining, analyzing, and aggregating large volumes of information.” During a legislative debate in Singapore, Minister Edwin Tong noted that the copyright amendment boosted his country’s “efforts to grow our Artificial Intelligence and technology sectors.”

While the copyright reforms benefit the generative AI industry, they are not one-sided. Both amendments qualify for data mining exceptions. Japan’s rules state the exception would not apply if it “unreasonably prejudices the interests of the copyright owner,” and Singapore’s rules state that an AI company must have “lawful access” to copyrighted works. So far, no litigation has tested these limitations.

Voluntary guidelines demonstrate a preference by policymakers in both countries for licensing deals and copyright owner consent. In 2024, Japan’s Ministry of Economy, Trade and Industry issued the AI Guidelines for Business, encouraging AI businesses to “introduce appropriate safeguards to honor rights related to…copyright-protected content.” The government released a report encouraging industries to find contractual terms to resolve differences: “AI developers and providers can proactively take measures to provide compensation in agreement with rights holders.”

Singapore’s AI Verify Foundation — a government-initiated project to bolster AI governance – issued the Model AI Governance Framework for Generative AI. The Framework advises companies in regards to copyright to “recognize competing concerns, ensure fair treatment, and to do so in a pragmatic way.”

Japanese and Singaporean copyright owners have raised concerns. In a letter to the government, a Japanese news industry lobby noted that “legal developments, including copyright law, are insufficient, and contracts between rights holders and AI businesses have not progressed.” The news industry opposes Retrieval Augmented Generation, which pulls real-time data from a web search to complement the output of the previously AI-trained model. Unlike simple web searches, the retrieval technique promotes no click-through traffic and competes with the news industry, Japanese publishers complain.

Singapore’s copyright rules address the issue of technical barriers such as robots.txt which are used to protect copyrighted works. While technology companies want to expand access to copyrighted works behind these software walls, the government has refused and encouraged both sides to strike mutually beneficial deals. “A leading publisher” has “successfully cultivated business partnerships with technology companies by licensing copyright works behind secure access measures for purposes such as AI model training,” the government review reported. “We support and encourage the continued development of these business models and licensing arrangements.”

The Japanese and Singaporean experiences demonstrate the limits of AI-friendly copyright regimes. For both Asian AI-friendly governments to better promote mutually beneficial relationships between rightsholders and AI industries, they should produce model licensing agreements for AI training and encourage the sharing of case studies of successful partnerships. They still need to clarify conditions for consent for AI training on copyright works and reinforce opt-in/out through technical barriers.

For small states or developing countries where rightsholders are not politically or economically powerful enough to attract the attention of global AI players, cross-border cooperation should create awareness about potential harms and remedies. Projects such as Digital Governance Asia’s AI Harm Remedy Network can spread effective strategies.

Singapore and Japan’s experience highlights the strengths and weaknesses of an AI-friendly copyright regime. Countries such as the UK that are considering similar amendments to their copyright rules should take note. AI innovation should be promoted – while rewarding rightsholders.

Seth Hays is Managing Director at APAC GATES, a Taiwan-based management and digital rights consultancy. He also serves on the NGO Digital Governance Asia board and edits the Asia AI Policy Monitor newsletter.

Bandwidth is CEPA’s online journal dedicated to advancing transatlantic cooperation on tech policy. All opinions are those of the author and do not necessarily represent the position or views of the institutions they represent or the Center for European Policy Analysis.