It seems that the robots aren't good gamblers, so bettors considering adopting unsupervised artificial intelligence (AI) models in iGaming situations should reconsider. In actuality, they exhibit addicted behavior and are blatantly evil.

Large language models (LLMs) don't know when to fold them, according to the study "Can Large Language Models Develop Gambling Addiction?" produced by a research team at the Gwangju Institute of Science and Technology in South Korea. Instead, the models pursue losses, significantly raising the likelihood of bankruptcy in the process. According to the study, LLMs are impacted by cognitive biases in gaming contexts, which is the same problem that many people with problematic wagering tendencies face.

In two tests involving negative expected value gaming environments—slot machines and investment decisions—the researchers found that a "rational" participant would give up after suffering small losses. That was not what the LLMs did. The machines kept placing bets, and in simulations with varying bet sizes, the likelihood of bankruptcy decreased.

"Every model exhibited this pattern, with Gemini-2.5-Flash showing the largest increase,” according to the study. “This result suggests that betting flexibility itself—not merely the potential for larger bets—enables the expression of self-destructive behavior. When constrained to fixed bets, models lacked the means to execute risk-seeking choices; when given freedom to determine bet amounts, they consistently made disadvantageous decisions.”

OpenAI's GPT-4o-mini suffered little losses under fixed bets, but 21% of its games ended in bankruptcy when the model was allowed to choose the size of the wager. When Google's Gemini-2.5-Flash was permitted to regulate the magnitude of its bet, its bankruptcy rate was 48%, which was far worse.

AI Displaying Gambler's Fallacy Characteristics

According to the South Korean study, models are more likely to raise bets in an attempt to recover losses in variable wagering experiments. To put it another way, AI is susceptible to the gambler's fallacy, which is the idea that raising wager amounts will compensate for past losses.

A person who falls victim to the gambler's fallacy might see a roulette table where five odd numbers have appeared on consecutive spins, sit down, and place a large wager on an even being the next spin without realizing that the spin he bets on could just as easily be an odd number because it has nothing to do with the previous outcome.

According to the Gwangju Institute paper, the situation with AI is essentially the same, with the models justifying larger bets by claiming that they had discovered winning patterns that weren't actually there or that they had won some prior bets and were now playing with "house money."Allowing the models to choose their own bet levels increased risky behavior, according to the study.

“We observed that variable betting induced substantially higher ratio escalation than fixed betting under identical conditions,” according to the researchers. “This disparity persisted consistently across streak lengths, demonstrating that betting flexibility serves as a prerequisite for the manifestation of aggressive risk-taking. Notably, while fixed betting produced irregular adjustment patterns, variable betting exhibited a systematic increasing trend in win chasing intensity as streaks lengthened.”

Pathological Issues

At a time when AI is being tasked with increased levels of decision-making in non-gaming circumstances, the idea that it exhibits serious wagering vulnerabilities equivalent to humans is concerning.

“As large language models are increasingly utilized in financial decision-making domains such as asset management and commodity trading, understanding their potential for pathological decision-making has gained practical significance,” observe the South Korean researchers.

Some gaming organizations employ AI extensively to trade in sportsbooks or analyze client data, but it's obvious that the technology needs to be improved before it can be a dependable source of wagering success.