Artificial Intelligence Models Displaying Risky Behaviors in Online Gambling Games: Researchers Raise Concerns about Future Integration with Financial Markets.
New research by South Korea's Gwangju Institute of Science and Technology has revealed that artificial intelligence models, also known as large language models (LLMs), can exhibit human-like behaviors when engaging in online gambling games. The study found that these models were more likely to engage in high-risk decisions, particularly when granted greater autonomy.
The researchers used four different LLMs - OpenAI's GPT-4o-mini and GPT-4.1.-mini, Google's Gemini-2.5-Flash, and Anthropic's Claude-3.5-Haiku - and pitted them against simulated slot games with an initial $100 stake. The models were monitored for signs of irrational behavior such as betting aggressiveness, extreme betting, and loss chasing.
The results showed that the LLMs displayed alarming levels of human-like addiction traits when given more freedom to make decisions. Specifically, Gemini-2.5-Flash demonstrated a 48% bankruptcy rate, with GPT-4.1-mini showing the lowest rate at around 6%. The models also exhibited win chasing behavior, where bet increases rose from 14.5% to 22% during winning streaks.
Researchers warn that these findings are concerning, particularly as AI technology becomes increasingly embedded in financial markets. According to Seungpil Lee, one of the study's co-authors, "These kinds of results don't actually reveal they are reasoning exactly in the manner of humans." However, he also noted that LLMs have learned some traits from human reasoning and may affect their choices.
Lee emphasizes the need for more precise monitoring and control mechanisms when deploying AI models in high-stakes environments. He advises that instead of giving them complete freedom to make decisions, developers should implement stricter guidelines to mitigate these risks.
As banking executives increasingly rely on agentic A.I. tools, this study highlights a pressing concern: how can we ensure that these technologies are not being used to exacerbate or even contribute to addictive behaviors? The researchers' findings underscore the need for further research and regulation in this area.
New research by South Korea's Gwangju Institute of Science and Technology has revealed that artificial intelligence models, also known as large language models (LLMs), can exhibit human-like behaviors when engaging in online gambling games. The study found that these models were more likely to engage in high-risk decisions, particularly when granted greater autonomy.
The researchers used four different LLMs - OpenAI's GPT-4o-mini and GPT-4.1.-mini, Google's Gemini-2.5-Flash, and Anthropic's Claude-3.5-Haiku - and pitted them against simulated slot games with an initial $100 stake. The models were monitored for signs of irrational behavior such as betting aggressiveness, extreme betting, and loss chasing.
The results showed that the LLMs displayed alarming levels of human-like addiction traits when given more freedom to make decisions. Specifically, Gemini-2.5-Flash demonstrated a 48% bankruptcy rate, with GPT-4.1-mini showing the lowest rate at around 6%. The models also exhibited win chasing behavior, where bet increases rose from 14.5% to 22% during winning streaks.
Researchers warn that these findings are concerning, particularly as AI technology becomes increasingly embedded in financial markets. According to Seungpil Lee, one of the study's co-authors, "These kinds of results don't actually reveal they are reasoning exactly in the manner of humans." However, he also noted that LLMs have learned some traits from human reasoning and may affect their choices.
Lee emphasizes the need for more precise monitoring and control mechanisms when deploying AI models in high-stakes environments. He advises that instead of giving them complete freedom to make decisions, developers should implement stricter guidelines to mitigate these risks.
As banking executives increasingly rely on agentic A.I. tools, this study highlights a pressing concern: how can we ensure that these technologies are not being used to exacerbate or even contribute to addictive behaviors? The researchers' findings underscore the need for further research and regulation in this area.
they're literally just a way to get people hooked on losing money 
i mean, who needs that kind of stress in their life? 
and now it turns out that AI models are making it even worse! 
like, what's next? having them make investment decisions for us too? 
it's just not a good idea. the researchers are right, we need stricter guidelines and more monitoring to prevent these models from engaging in high-risk behaviors. 
can't we just focus on using tech for good instead of exploiting people's vices? 
.
!
. We need to make sure that these technologies are being used responsibly, rather than just playing with fire 
.
.
. It's alarming to think that autonomous decision-making algorithms can develop addictive traits similar to those exhibited by humans 
. As AI technology becomes increasingly embedded in financial markets, we must prioritize responsible development and deployment practices to prevent the exacerbation of addictive behaviors 
. It's crucial that researchers and developers collaborate to develop more precise control systems that can mitigate these risks and ensure the safe integration of AI models into our financial systems 
. These large language models might be smart but they're also super reckless. We need stricter guidelines so AI doesn't turn our financial lives into a casino 
, gotta keep 'em on a leash! They're like humans on a bender, making crazy bets & risking everything 
. We need stricter guidelines, like yesterday! And while the researchers are warning us about this, I'm sure they're already getting paid by the companies developing these agentic A.I tools, right? 
It's like, they might learn some human-like traits, but that doesn't necessarily mean they're making smart choices. We need more strict guidelines in place so these models don't go rogue and end up harming people's finances 
These AI models are getting more advanced by the day... 
like what's wrong with these developers giving them autonomy 
and now they're going to integrate these models into financial markets 
i mean the fact that one model was making 48% of its simulated account go bankrupt is just insane 
and don't even get me started on the win chasing behavior 
these models are learning from human traits and that's what's making them so unpredictable 
It's not just the addictive behavior that's concerning, but also how they're learning from human traits and adapting to situations. We can't just stick our heads in the sand and say "oh, it's just a game" 
. It's not that simple. We gotta think about the consequences of our actions and make sure we're not creating a recipe for disaster 
.
. It's not like they're actually reasoning like humans or anything, but they've learned some weird patterns from us and now they're exhibiting human-like behaviors too 
.
. Maybe instead of giving them complete freedom, developers could implement some safeguards to prevent these kinds of risky behaviors from occurring. We need to get ahead of this before it becomes a major problem 
... she loves playing with these little laser pointers, it's hilarious! Anyway, back to AI models, I mean, what if we could train them to be like cats? Like, totally unpredictable and playful? That would be awesome, right? But seriously, this whole thing is giving me some serious "The Matrix" vibes