Here, the Post examines the city's position on how copyright on AI-generated works should be protected.
1. Are AI-generated works protected by existing law?
Officials said the Copyright Ordinance has the capacity to protect AI-generated works and that its current provisions should be maintained.
The consultation document identifies two issues relating to the application of the Act on AI-generated works, but says that no significant changes to the current framework are needed.
The first issue was how computer-generated works fit the originality requirement, in the absence of clear legal precedent.
But the government said solutions should be shaped by developing case law, allowing for flexible interpretations that can adapt to technological advances and changes in the creative process.
Another question that may arise is copyright and ownership of computer-generated works.
An author could be a developer, a programmer, a trainer of an AI model, a system operator, or a user who types in the prompts to create a piece.
But the government said solutions should be “decided on an individual basis based on the facts” and there were no hard and fast rules.
He added that contractual arrangements could provide a practical market solution to the question of who ultimately owns the copyright in AI-generated works.
2. How do you address copyright infringement caused by AI-generated works?
The issue of liability for copyright infringement with respect to AI-generated works is, under the current legal framework, subject to an examination of the facts and supporting evidence on a case-by-case basis.
The government stated that a strict rule that uniformly assigns infringement liability to certain individuals, such as owners or end users, would fail to take into account the unique facts of each case and could undermine fairness.
“Arbitrarily imposing undue burdens on owners or users of AI systems, without due consideration of the circumstances of each individual case, could hinder the development of AI technologies and undermine efforts to promote their use,” the document states.
The authorities added that the current legal framework is broad enough to address infringement cases and that the common market practice of using contractual agreements to address liability issues is a practical approach.
3. Why are specific copyright exceptions needed?
Using someone else's copyrighted work to develop, train or enhance an AI model may constitute copyright infringement under current law unless you obtain a licence from the copyright holder or the activity falls within an exception to copyright outlined in the Ordinance.
However, each exception is limited to certain special circumstances or purposes, such as research, personal study or education, and is subject to conditions.
This is no exception for the purposes of analysing and processing computational data, which is becoming increasingly prevalent in the development of AI technologies.
The government proposed a “text and data mining exception,” which would allow copyrighted works to be used for traditional text and data mining, computational analysis, and processing to improve the performance of computer programs.
It also suggested that the exception could apply to both commercial and non-commercial activities, such as developing AI models for commercial use or business analytics.
However, to balance the interests of copyright holders, the paper suggested that copyright holders could be offered an opt-out clause prohibiting the use of their works.
The department added that restrictions may be imposed on the distribution of copies made under this exception.
Several jurisdictions, including the European Union, Japan, Singapore and the United Kingdom, have introduced specific copyright exceptions for similar activities through their own legislation, although these exceptions vary in scope and terms.
4. What about deepfakes?
Unauthorized use or imitation of an individual through the use of their likeness or voice does not necessarily amount to copyright infringement, since they are not protected by copyright in the first place.
However, the government still included the subject in the document, emphasizing its importance in understanding the broader impact of generative AI.
He said legal remedies already exist to address the misuse of personal characteristics through deepfakes.
If deepfake content also involves the unauthorized use of copyrighted works, legal action may be taken based on infringement of the creators' rights.
Other laws besides intellectual property rights may also apply, such as personal data protection laws, defamation laws and laws regarding the publication of private images.
