AI GambitHunter: How It Hunts Online Gambling Sites in Indonesia

Jakarta – The proliferation of online gambling in Indonesia continues to pose a significant challenge, with thousands of sites blocked only to be replaced by new platforms employing evolving transaction methods. Traditional blocking strategies are proving insufficient, prompting a search for more sophisticated solutions. Now, a team of three Indonesian engineers has developed GambitHunter, an artificial intelligence (AI) system designed to identify online gambling sites and trace the financial transactions used by operators.

The rise of online gambling has become a serious issue in Indonesia. According to data from the Indonesian Ministry of Communication and Digital Affairs, approximately 4 million Indonesians were involved in online gambling as of October 2024. The Jakarta provincial government has responded by offering consultation services to residents seeking help with problem gambling, recognizing the broader social and economic impacts beyond mental health. The Financial Transaction Reports and Analysis Center (PPATK) has also reported a substantial surge in online gambling transactions, reaching over Rp600 trillion in the first quarter of 2024, following Rp327 trillion in 2023. The Indonesian government has declared a “total war” against online gambling and money laundering, strengthening national coordination across agencies.

GambitHunter, created by Steven Sukma Limanus, Ilham Firdausi Putra, and Reynaldo Wijaya Hendry, aims to address this escalating problem. The system’s development was remarkably swift, winning second place in the OpenAI Codex Hackathon in Singapore with a prototype built in just seven hours. “I feel this will become the new normal because AI really increases developer productivity,” said Steven, highlighting the efficiency gains enabled by tools like Codex.

AI Agent Exploration

Unlike manual investigations, which can be lengthy and resource-intensive, GambitHunter employs an AI agent to automatically explore the internet, searching for and analyzing sites suspected of hosting online gambling activities. This agent proactively seeks out potential platforms from various online sources. “In the exploration phase, we try to find online gambling sites from several sources. There’s an AI agent that does this automatically,” explained Steven.

The system then compiles a list of potentially suspicious sites for further analysis.

Classification Using AI Models

Once a list of sites is gathered, GambitHunter utilizes AI models to classify whether a site is genuinely involved in online gambling. Reynaldo Wijaya Hendry explained that the system leverages large language models to aid in this identification process. “We use AI for this. So the exploration agent will try to collect as many potential gambling sites as possible. After that, we use the OpenAI GPT-5.2 model to classify whether the website is actually an online gambling site,” Rey stated. This approach allows the system to process a large volume of sites concurrently, eliminating the need for individual manual review.

Accessing Systems with Dummy Accounts

If a site is flagged as a potential online gambling platform, GambitHunter proceeds to investigate further by accessing internal sections of the site. To gain access to restricted areas, the system can automatically create dummy accounts, allowing the AI to view features typically reserved for registered users, such as deposit pages and payment methods. This provides insight into the transaction flow within the site.

Extracting Payment Data

After understanding the transaction structure, GambitHunter enters the data extraction phase, attempting to identify payment information displayed on deposit pages. This includes bank account numbers, phone numbers, and images or screenshots of payment pages. “Currently, we can extract images and account numbers used to receive deposits,” Rey noted. He added that the system has the technical potential to interact automatically with online gambling site customer service through chat functions. “We haven’t had a chance to test this directly, but technically I’m quite confident that it’s also possible because there are no fundamental technical obstacles,” he said.

This automated process significantly reduces the time required for investigations, which would be considerably longer if conducted manually.

AI’s Role in Development

The development team emphasized that AI technology was integral not only to the functionality of GambitHunter but also to its creation. Ilham Firdausi Putra stated that the OpenAI Codex model played a crucial role in designing the architecture and writing the code. “From designing the architecture, aesthetics, to the program code, everything used Codex. I was surprised by how effective and capable Codex is,” Ilham said. The hackathon organizers even provided participants with USD 200 worth of Codex subscriptions to maximize its use during the competition.

Potential for Law Enforcement Assistance

The developers hope GambitHunter can be utilized by law enforcement agencies to map the networks behind online gambling, which have historically been difficult to trace. Steven stated that the project’s primary focus is to assist stakeholders in maintaining a secure digital ecosystem by identifying transaction routes for online gambling. “Our focus is to help stakeholders maintain the digital ecosystem by finding and identifying online gambling transaction routes,” Steven explained.

Ilham added that the team believes systems like GambitHunter are best operated by government agencies with law enforcement authority. “We agreed from the beginning that GambitHunter should ideally be operated by government agencies with authority. We are also open to sharing code and knowledge with parties interested in this issue,” Ilham said. During the hackathon demonstration, the system successfully identified real online gambling sites and extracted data from them. “The video we showed was the result that GambitHunter found directly at that time. It wasn’t a dummy site, but a real site found by the system,” Ilham clarified.

For the development team, GambitHunter demonstrates how AI technology can be leveraged to understand patterns of illegal activity on the internet. With faster and more systematic analysis, law enforcement against online gambling is expected to become more effective in the future.

As of November 2024, Jakarta recorded the highest number of online gambling cases in Indonesia, with 1,856 children exposed and transactions totaling Rp2.29 billion, according to Penjabat (Pj) Governor Teguh Setyabudi. This data, sourced from PPATK, highlights the urgent need for tools like GambitHunter to combat the growing problem and protect vulnerable populations.