The New York Times v OpenAI
In December 2023, the New York Times (NYT) launched a lawsuit against Microsoft and OpenAI, claiming that OpenAI used NYT articles to train its GPT large language models without permission, violating US copyright and competition law. However, proving their claim would require NYT to have access to OpenAI’s source code—highly confidential and valuable data not usually disclosed, especially by software-based companies. After extensive negotiations, a US magistrate has issued an order granting NYT access to view OpenAI’s source code under strict protocols and conditions.
Although NYT’s lawyers can now view the source code, they only must do so in a secure inspection room and on a secured computer not connected to the internet. Additionally, no recordable media devices, including computers or phones, are allowed in the room. NYT’s lawyers are permitted to take handwritten notes or electronic notes on a provided note-taking computer in scratch files, but they cannot copy any of the data itself.
Comparison with Australian courts
While both US and Australian courts permit access to source code, Australian courts are generally cautious in requiring extensive disclosure of proprietary source code where not necessary. Similar to the US, Australian courts can also impose restrictions on that disclosure. For example, cases such as Alcoa of Australia Ltd v Apache Energy Ltd illustrate how Australian courts can establish confidentiality regimes for the inspection of discovered documents, focusing on the need to protect commercially sensitive information from unnecessary exposure.
Similarly, in Motorola Solutions, Inc. v Hytera Communications Corporation Ltd (Liability)1, the Court acknowledged that Motorola’s source code constituted information of ongoing commercial value, warranting the maintenance of its confidentiality. As a result, the Court was prepared to protect that commercial value by imposing a confidentiality regime governing access to those documents; restricting access to:
- Motorola;
- its legal representatives; and
- experts in the proceeding.
Specifically, when requesting preliminary discovery of source code or otherwise, the court must be satisfied that a prospective applicant:
- reasonably believes it has the right to obtain relief in the court
- after making reasonable inquiries, does not have sufficient information to decide whether to start a proceeding in the court to obtain that relief
- reasonably believes that a prospective respondent has in their control the relevant documents and inspection of those documents would assist in making the decision as to whether to obtain relief.
Aristocrat Technologies Australia Pty Ltd v Light & Wonder, Inc.
A recent example – Aristocrat Technologies Australia Pty Ltd, a gambling machine designer and manufacturer, was awarded preliminary discovery against Light & Wonder, Inc.
Aristocrat’s application for preliminary discovery was granted in respect of mathematical rules, formulae and models to allow Aristocrat to assess whether Light & Wonder’s “Dragon Train” game was copied from Aristocrat’s “Lightning Link” and “Dragon Link” games. However, the judge declined to require the production of the source code itself, as “there did not appear to be a reasonable basis to believe” that Aristocrat’s source code had been copied. The only evidence of copying related to those rules, formulae and models.
Key takeaways
The handling of source code disclosure in legal disputes varies not only between jurisdictions but also on a case-by-case basis. The ongoing lawsuit between the NYT and OpenAI highlights that Courts are recognising the value of proprietary source code to companies which primarily deal in software and AI; imposing discovery conditions which grant the bare minimum level of source code access which is necessary to achieve the purpose for which discovery has been sought.
It remains to be seen whether, in a case seeking discovery of source code in Australia, the Australian courts would impose a similar regime to that imposed in the NYT / OpenAI case. However, as a critical threshold issue, in order for a technology company to defend a discovery application relating to proprietary source code access, that company will likely need to show that it has stringent internal confidentiality and access regimes in place which would be undermined by the grant of broad discovery rights to a third party; that is, that the source code is not already broadly accessible, and that limited discovery conditions are necessary for the company to preserve the value of its proprietary software.
For more information, please contact Sarah Gilkes.
1[2022] FCA 1585
