Australia has made significant progress in its fight to wrest some revenue back for news from social media platforms and search engines, but experts caution against rushing similar efforts for generative AI. 

The News Bargaining Incentive (NBI) cleared parliament last month, paving the way for Australian news publishers to resume receiving compensation for their content from tech giants such as Meta and Google. 

But the notable absence of generative AI from the legislation’s scope represents a “missed opportunity”, RMIT School of Media and Communications associate professor James Meese tells PIJI. 

He says generative AI presents a greater threat than traditional digital platforms, likening the technology as more akin to the advent of the internet than social media. 

While news publishers choose to share content on social media platforms and link back to their own brands and websites, generative AI takes and presents news content itself through platforms like ChatGPT or features such as Google’s AI Overviews, he says. The latter appears to be particularly damaging to click-through rates.  

The Australian government has signalled its intent to address AI’s impact on news in combination with its impact on other sectors through copyright reforms set to be revealed next year.  

Details about what these reforms could look like are scant, but the government ruled out giving AI companies a text and data mining exemption. 

Although there may be some frustration in the news industry at the lack of progress on AI, Meese says the government is moving much faster than it has on previous digital change. 

He says tackling the complex issues presented by AI should take some time to get the best result. 

“I think the government, quite rightly, is looking at [generative AI] as a kind of epochal change,” Meese says. 

“So it might mean that while specific sectors feel that things aren’t moving fast enough – and I myself am noting that there was an opportunity here to do something with respect to the [NBI] – there’s an argument to be made that you do it once and you do it right.” 

Although generative AI has become pervasive since the late-2022 launch of OpenAI’s ChatGPT, Macquarie University associate professor in law Rita Matulionyte says the government’s current timeline to present new copyright legislation in early 2027 is ‘unrealistically’ fast. 

“You could say three years in waiting and thinking about what to do [is too long],” she says. 

“But I think we should acknowledge that no other country has found a solution, because technology keeps evolving, and with each new [evolution], you get each new aspect of technology which we haven’t even thought about.” 

She points to the EU’s 2019 Directive on Copyright in the Digital Single Market, which provided copyright exceptions for text and data, permitting users and entities to mine text and data without remunerating rights holders as long as the content was lawfully accessed and rights holders had not expressly reserved their rights. 

The socio-economic consequences of that decision transformed significantly following the generative AI boom that came just a couple of years after the directive was adopted, Matulionyte says. 

“Technology changes so fast that that we have to be very careful not to create more problems [from] solutions.” 

What could copyright solutions look like? 

Currently, the Australian government appears focused on using copyright reform to tackle the impact of generative AI on creative industries and media. 

Matulionyte says there are several models that could be adopted, which should address both past and future harm caused by generative AI. 

She says one reform option could be the introduction of ‘compulsory licensing’, similar to the statutory licence Australian educational institutions operate under.  

Under this scheme, educational institutions pay fees for certain uses of copyright material to ‘collecting societies’ such as the Copyright Agency. The fees are then distributed to rights holders, who cannot opt out of the scheme. 

Then there is ‘extended collective licensing’, which Matulionyte says is not yet common in Australia. This option allows collecting societies to negotiate agreements for use of content covering both members and non-members in the same field, while allowing rights holders to opt out. 

She says an option that sits outside copyright law is a cultural fund into which AI companies would pay fees, which would then be distributed through the news industry by a government or independent body. 

Some combination of these options may work, she says, as long as there is space for direct agreements for publishers able make such deals, opt out options so companies or individual journalists can retain their autonomy, and equitable distribution of funds between stakeholders. 

Issues copyright reform may not address 

Aside from the use of content for AI training and content creation purposes, Meese says there also needs to be focus on supporting media diversity, both for the sustainability of small publishers and for audiences to access a wide range of news sources. 

Major publishers such as the Guardian and News Corp have already struck licensing deals with OpenAI, allowing their content to be used in search results. 

In Australia, Nine Entertainment signed a deal with Microsoft in July, formally allowing Microsoft Copilot to access and reference Nine publications’ content during AI searches. The arrangement reportedly means Copilot will display snippets, headlines and summaries, but users will be directed back to original websites for the full articles. 

But few deals have been made between AI companies and smaller publishers. In this case, collective action may be most viable, as seen in the deal announced in March between AI startup Bria and the US-based News/Media Alliance, which represents more than 2000 organisations. 

“The value derived from Australian journalism needs to support a diverse news ecosystem rather than just a few specific partners,” Meese says. 

“If there are licensing deals between two or three publishers, does that mean their content gets presented over a range of other content provided? Obviously … it’s not that simple, but those are the kind of questions that may not be fully answered through the prism of copyright.” 

Sezen Bakan