Hong Kong’s government is considering a major rewrite of its copyright laws to make the region more attractive to AI companies.
In a white paper published by Hong Kong’s Intellectual Property Department to kick off a two month public consultation, the government says that it would be “justifiable” to make changes to copyright rules to “further enhance Hong Kong’s copyright regime to ensure that it encourages creation and investment in creativity while supporting innovation”. Hong Kong has its own system of copyright law that is separate to that of China.
The consultation, which was announced on 8 Jul and runs through until 8 Sep, explores the introduction of “a new and specific copyright exception to allow reasonable use of copyright works for computational data analysis and processing, covering conventional text and data mining and the training of AI models”.
This would help “foster the growth of the AI industry” and allow AI companies to include copyright protected works in their training datasets without getting permission from copyright owners. The report goes on to say that introducing a data mining exception would help develop “an international innovation and technology centre which lies at the heart of sustainable economic growth of Hong Kong”.
The government presumably knows that the new exception will be opposed by copyright owners. However, it says that the “overall benefits” that could result, including“driving and boosting the development of AI technology”, are justification for making the change, and that any drawbacks could be “balanced out by providing appropriate safeguards to copyright owners”.
Those safeguards may include an opt-out that allows copyright owners to exclude specific works from being used, with a proposal to introduce a provision that renders “text and data mining activities unauthorised if licensing schemes are available or copyright owners have expressly reserved their rights”.
The consultation in Hong Kong comes as the music industry continues to lobby law-makers around the world with the aim of clarifying the copyright obligations of AI companies.
For the music industry, and other copyright sectors, what those copyright obligations should be is clear and unambiguous: any company using copyright protected content to train its AI should have to get explicit permission from relevant copyright owners, and should not be able to rely on data mining exceptions or the broader US principle of fair use.
Both Singapore and Japan currently have data mining exceptions in copyright law that can be relied upon by AI companies that are based there.
In the EU, the 2019 Copyright Directive also introduced a data mining exception, however, that comes with an opt-out which allows copyright owners to ‘reserve’ their rights. Where copyright owners do that - and many have already - an AI company still needs to get permission from the copyright owner to use their content. The opt-out proposed in the Hong Kong white paper could work in a similar way.
UK copyright law has a data mining exception for non-commercial research. In 2022, the UK government proposed introducing a wider exception that could be relied upon by commercial entities, but that plan was abandoned following a major pushback from the copyright industries.
The Hong Kong government is clear that its exception should apply to commercial entities too. It writes, “to optimise” the proposed exception and “realise its benefits for a wider sector of the public, we suggest that the exception should not be restricted to non-commercial research and study”.
For countries keen to become commercial hubs in the global AI industry, it is tempting to provide protection for AI companies that need access to large quantities of content to train their models, making it attractive for AI businesses to set up shop.
The music and other copyright industries know that this means it is likely there will always be a small number of countries where AI companies can rely on data mining exceptions.
This presents a significant challenge as, in theory, it means that an AI company could circumvent claims of copyright infringement by locating their servers and training datasets in countries that provide a favourable legal framework, and then make their AI products based on that training available globally.
To that end, lobbyists working for the copyright industries are also campaigning for market access restrictions. In simple terms this would mean that, if you develop an AI model and train it with copyright protected data in a country where copyright exceptions mean that is allowed, then the model can’t be made available in other countries where an explicit licence would have been required for the training data.
In the recent report on AI published by the UK All Party Parliamentary Group For Music, this was voiced in a recommendation that suggested “as a condition of market access, the government should require AI models to comply with UK copyright provisions, even if the services or goods they have developed are created in compliance with the laws outside the UK”.
Where such market access restrictions do go into force, it will be interesting to see how they are enforced - or even if such enforcement is possible.
While the current global technology companies will ultimately need to be compliant, new companies based in other countries may take a different approach, and with the current rate of development in AI it’s entirely possible that in a few years’ time today’s current tech giants will be dwarfed by companies that are unknown today.
One outcome might be another round of web-blocking by copyright owners, seeking injunctions that force internet service providers to block access to those AI platforms.
Generally ISPs are very compliant with such injunctions these days. However, web-blocking has limited effectiveness, and the copyright industries might end up in another game of whack-a-mole trying to stay ahead of non-compliant AI platforms.