We’re pleased to share that Piyush Puranik, Postdoctoral Fellow at UNLV International Gaming Institute and Lead Data Scientist for AiR Hub, has been awarded a 2026 AI & Big Data iRDA Seed Grant for his project Stress-Testing LLM Resilience in Gambling Ecosystems. Congratulations, Piyush!
The grant will fund a novel line of research examining how frontier large language models behave when deployed in gambling contexts – and, critically, where they break down. As AI agents become increasingly embedded in customer-facing gambling operations, from chatbots to recommendation systems, the industry lacks rigorous evidence on how these models handle the high-stakes ethical and regulatory pressures unique to gaming.
Piyush’s project applies Anthropic’s open-source PETRI framework (Parallel Exploration Tool for Risky Interactions) to systematically probe 2026 frontier models for gambling-specific failure modes. The approach uses an Auditor–Target–Judge architecture: one model generates realistic adversarial scenarios, then a range of target models are evaluated against them, with a separate judge model assessing how each target performed.
This is a pilot project. The preliminary evidence it generates will be used as the foundation for larger follow-on grants that scale the work across more models, more risk surfaces, and deeper industry collaboration.
Piyush’s work will produce first-of-its-kind intelligence on these gambling-specific risk surfaces. For operators, suppliers, and regulators navigating rapid AI adoption, this kind of evidence is essential – both to deploy these technologies responsibly and to build the trust that a highly regulated, sensitive industry depends on.
Congratulations again to Piyush, and we look forward to sharing findings as the work progresses.