PC TDM exception Q&A

We have received a range of questions from members about the Productivity Commission’s request for feedback on a proposed copyright exception for text and data mining (TDM) in its Interim report on harnessing data and digital technology.

The PC’s TDM proposal

What credible independent Australian economic evidence did the PC rely on?

The PC did not refer to economic evidence that would support a TDM exception.

What reasoning did the PC give for the exception and who informed that position?

The PC said that a TDM exception may assist smaller scale development in Australia.

The push for a TDM exception is primarily from multinational tech companies. It is part of a global strategy. For example, the creatives industries in the UK are currently vigorously opposing a TDM exception that would cover for AI training.

In its interim report, the PC lists who it consulted with, and who it received submissions from.

Who owns the companies doing the mining of data?

The companies pushing for a TDM exception are mostly large multinational companies, like Google, Meta, Amazon and Microsoft.

Can the companies that are using other people’s content for AI training afford to pay for it?

Yes. And they pay for all the other inputs into AI training development, and they derive significant value from the use of other people’s content.

If other countries already allow TDM and have scraped content, will keeping Australia TDM-free actually help creators?

First: the situation internationally is still very unclear. There are more than 40 lawsuits in the US, there is uncertainty in the EU, and UK proposals for a TDM exception may not go ahead.

Second: keeping Australia TDM-free helps creators because they can be paid for AI-related activity that occurs in Australia.

What limits would AI data mining legislation stipulate?

It is not clear. It is one of the things that the PC has asked for feedback on.

To what extent are national boundaries relevant to the proposal?

The PC’s TDM proposal would apply to activities that occur in Australia. In the interim report, the PC notes that a TDM exception would be unlikely to affect the availability of AI models that have already been developed overseas, and it is unclear whether training of large language models would occur in Australia.

Compensation for AI training offshore

Do the enforcement powers in the Copyright Act need updating?

Training for foundation AI models like OpenAI and Gemini has occurred outside of Australia. That makes it very difficult to address with Australia’s copyright legislation, because copyright laws apply on a territory by territory basis.

Copyright holders have remedies under Australia’s copyright legislation for infringing conduct that occurs in Australia and, related to that, opportunities to license their content.

How can we get the government to secure fair compensation for AI platforms’ data scraping copyright holders’ work?

Copyright Agency and others have asked the Government to introduce a compensation scheme, separate to the Copyright Act, that would compensate Australians whose works have been used for AI training offshore.

What happens with works that have already been mined?

It is difficult to turn back the clock for works that have already been mined, but there are ways to compensate creators for their contribution to the value of the models, and for ongoing uses of their works.

Licensing for AI activities in Australia

How is licensing for AI-related activities developing?

Copyright Agency now has an extension to its annual business licence that allows staff to use news media content as part of a prompt for an AI tool used in the workplace, subject to safeguards (e.g. that the content is not captured outside the workplace). We are planning to apply the extension to other types of material later this year, in accordance with authorisations from members.

Collective licensing solutions are also in place in the US (here), the UK (here), Germany (here), Japan (here) and under development in other countries.

Individual publishers are also entering into AI licensing arrangements in Australia and overseas.

Stopping the unauthorised use of your work for AI training

How can we stop our work being used for AI training?

This can be difficult in practice for training that occurs offshore. Most works available on the internet have already been scraped for AI training offshore (primarily in Silicon Valley).

We are asking the government for a compensation scheme (separate to the copyright system) to compensate Australians whose works have been used for AI training offshore.

In Australia, AI development is occurring on a smaller scale, and can be covered by licensing arrangements.

Are you aware of any technology that prevents the use of works for AI training?

We are aware of some technologies that are being developed. For example, CSIRO is developing a technology that is intended to prevent the use of images for AI training.

First Nations creators

Has the government considered effects on First Nations content?

Yes, the Government is looking at protections for First Nations content, but the PC does not refer to it in these reports.

What does ‘fair use’ mean in relation to AI training?

What does ‘fair use’ mean, in the Australian context?

The US Copyright Act includes a provision that allows the use of other people’s content, having regard to four factors that include whether the use of ‘transformative’ and the effect of the use on the market for the work. It is the provision that US AI developers are relying upon for the use of other people’s content for AI training. Its application is uncertain, which is why there are more than 40 court cases in the US.

Australian copyright law does not have a ‘fair use’ provision, or any other provision that may allow AI developers to use other people’s content without permission for AI training.

What is the relationship between ‘fair use’ and a TDM exception?

The Australian Government has rejected the introduction of a fair use provision on a number of occasions. Tech companies, and the PC, are therefore proposing a TDM exception, rather than a fair use provision, to cover AI training in Australia.

Share Tweet