Taiwan’s government is exploring how AI developers could pay for material used to train their models, while building a voluntary corpus of traditional Chinese content for a sovereign AI project. The proposal could shape how creators and AI companies negotiate over training data, but the payment mechanism is still only a question under study.
Watch Desk analysis
What happened
The Taipei Times reports that the Ministry of Digital Affairs is collecting four kinds of material for its corpus: public-domain works, book descriptions provided by publishers, book previews and works authors have voluntarily licensed. The ministry says it has not sought training material from Taiwanese news outlets.
The corpus had reached 3.3 billion tokens by the middle of October, according to the report. The ministry has not said how any payments to content owners would be calculated, how revenue might be shared or when a payment system could arrive. Its stated aim is for large AI companies, particularly those based in the US, to pay for content used to train their models.
Why it matters
This is a live policy question with consequences for both AI developers and the people whose work may be used in training. Taiwan’s current approach emphasises public-domain and voluntarily licensed material, while the government says it is also considering ways to encourage payment. That leaves the practical details, and the boundaries of any eventual scheme, unresolved.
The report also captures a real disagreement: some media representatives want a licensing platform and disclosure of licensed works and training-data sources, while an industry association representative argues that fair-use provisions should remain available in some circumstances. Those are competing proposals, not settled rules.
Our read
The interesting news is the direction of travel, not a payment scheme that does not yet exist. Taiwan is assembling a defined pool of content for its own AI effort and studying how to connect rights holders with developers. The next meaningful signal will be whether the government turns that broad ambition into terms people can actually accept or challenge.
What to watch
- Whether Taiwan proposes a payment formula, licensing platform or collective mechanism.
- What content the government adds to the corpus, and on what terms.
- Whether the rules for publisher participation and author opt-outs change.
- Whether the debate resolves how licensing proposals interact with fair-use arguments.
Discussion spark: Should AI developers have to pay for training material by default, or should public-domain access and fair use remain the starting point?
Sources and evidence
- Government studying ways to compensate creators whose work is used to train AI – Taipei Times (27 October 2026, 16:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.