The gambling industry is no stranger to lawsuits. Between Kalshi and a growing majority of US gambling regulators, or the odd unhappy customer suing operators for potential losses, there is little time for lawyers to rest in this sector. However, when Aristocrat accused Light & Wonder (L&W) of using trade secrets and protected works in their new slot game, this introduced a topic that is not quite as common in the sector: copyright infringement.

When L&W released its Dragon Train game in March 2024, there were immediate comparisons with Aristocrat’s Dragon Link series, which was first launched in 2017. Upon closer inspection, Aristocrat found that several of its former executives had left the company for senior roles at L&W. Many of these individuals would have had access to intellectual property, including the code used for game mechanics and features. In the settlement, L&W acknowledged that it had used “certain Aristocrat math information” in the development of both Dragon Train and Jewel of the Dragon and that it would “permanently destroy all documents reflecting that information.” L&W promised never to use the math again, and that it would immediately cease the sale of these games worldwide and remove existing installations. All of this would come with a hefty $127.5m compensation price tag from L&W, too, in respect of the claims for misappropriation and infringement of Aristocrat’s intellectual property.

Now, while the full details have not been released to the public, responsibility has been accepted. But with the rate at which executives are pushing AI into the work process, how long might it be until we see this same story, but through artificial agents copying a competitor’s work instead?

How involved is AI within game development?

Game development teams are no strangers to AI. Whether it is using Qwen2.5 72b to generate long-form code or assist in speculative decoding, or perhaps using Midjourney to create graphics, developers are either being warned against using them – or being encouraged to – depending on the studio they work at. After all, attitudes against AI are in a major shift at the moment.

AI models are hailed as transformative for the development process, speeding up production time and streamlining different processes. However, as time has gone on, there is certainly some anti-AI sentiment out there. In the traditional video game industry, Clair Obscur: Expedition 33 was stripped of its “Game of the Year” and “Debut Game” achievements at the 2025 Indie Game Awards because generative AI was used during development for placeholders, even though these assets were replaced with human-made textures before release.Generally, as much as AI assists, there is definitely a feeling in some quarters that people want to engage with creative endeavours made by other people, supporting their careers and livelihoods in
the process.

But players do not seem to form the same emotional connections with casino games as they do video games. For example, you do not typically see people wearing Big Bass Bonanza hoodies in the same way they do with Pokémon, nor are people buying plushies from Fluffy’s Favourites. So it seems obtuse to assume players will hold gambling game developers to the same high standards, which means the industry is safe, right?

Is it a case of finders keepers? 

To generate anything, AI models need to scrape content created by people and ‘learn’ what users expect from their prompts. This has led to dozens of high-profile lawsuits, including accusations against Nvidia that the company tried to make a deal with the notorious pirated book databank Anna’s Archive for “high-speed access” to the library and train its LLMs on the stolen content. Similar legal action has been taken in the music industry against AI models and even in journalism, as the BBC threatened to take action against Perplexity AI after it was found generating stolen content “verbatim” from the news corporation.

This poses the question: with how prolifically AI is being used in game development, alongside how much content AI steals from other sources, how long could it be before an AI agent spits out code, math or other protected content from a competitor and lands the studio in trouble? Unless the employee can prove that they developed that particular math model or mechanics from scratch, there could be scope for legal action and accusations of stolen content.

While the exact process of scraping content has been kept pretty hush-hush (and understandably so, how would these LLM companies frame stealing copyrighted work to create profit for themselves in a positive light?), there is an easier way to collect information. Many employees are pasting highly sensitive and confidential data straight into these AI models, such as ChatGPT and Claude, and requesting them to ‘neaten aspects up’ or provide feedback on the work. This can pose some issues. Even if you opt out of settings designed to retain and remember your chat history, parent company OpenAI will still store any conversations within ChatGPT and review the transcripts for up to 30 days.

In January of this year, OpenAI landed itself in hot water when it asked contractors to upload work they had completed so engineers could compare the quality of AI output with that of documents from professionals. The contractors did not need to worry about OpenAI stealing confidential information either, as they were encouraged to use a tool created with ChatGPT that could identify this sort of data and remove it from the documents. That is right, you can remove any critical data from being seen by an AI model by using an AI model deliberately yourself beforehand.

Can aspects of AI pose a threat to the gambling industry?

Gambling has already created an interesting relationship with AI. Beyond game development, sportsbook operators have begun replacing traders with AI agents; in turn, players are creating AI models to analyze the market and place bets en masse. Within the space of a few months, prediction markets have already succumbed to a tidal wave of AI betting activity. It is no longer simply a fun way to predict the outcome of real-world events, but rather a competition to see whose AI model can predict these outcomes and place the most value-creating bets in the shortest span of time.

As for whether game studios need to be cautious about their AI models accidentally regurgitating work from their competitors, this remains a viable threat. Some companies in the gambling space are offering custom games created in just a few weeks through the assistance of AI; while others, like slotGPT, hand the entire process over to AI, which can “turn a simple prompt into a fully customized and playable slot game” in a few seconds. The joint statement from L&W and Aristocrat explained that both companies acknowledged the significant investment and innovation that went into game design and development, including the complex and confidential underlying math models that remain the backbone of casino titles.

In the words of Trevor Croker, Aristocrat CEO and Managing Director: “The court recognised that Light & Wonder was able to develop Dragon Train by using Aristocrat’s valuable trade secrets and without investing the equivalent time and money.” Although this case was the result of people’s actions, which seems ironic when framed within an argument against AI stealing from hard workers, the final point from Croker may still ring true.

Would Company A have a leg to stand on if its  AI model regurgitated the exact code that Company B developed, considering it had not invested any time or money into the process? There is a valid argument for both sides, and although LLM models and AI companies have fared well in courts so far regarding stealing copyrighted content, this could be a different matter when it comes to trade secrets rather than simply stolen artwork.